proposed criteria for the assessment of low frequency

Proposed criteria for the assessment oflow frequency noise disturbance

Moorhouse, AT, Waddington, DC and Adams, MD

Title Proposed criteria for the assessment of low frequency noise disturbance

Authors Moorhouse, AT, Waddington, DC and Adams, MD

Type Monograph

URL This version is available at: http://usir.salford.ac.uk/id/eprint/491/

Published Date 2005

USIR is a digital collection of the research output of the University of Salford. Where copyright permits, full text material held in the repository is made freely available online and can be read, downloaded and copied for non-commercial private study or research purposes. Please check the manuscript for any further copyright restrictions.

For more information, including our policy and submission procedure, pleasecontact the Repository Team at: [email protected].

mailto:[email protected]

Prepared for Defra by Dr. Andy Moorhouse, Dr. David Waddington, Dr. Mags Adams

Proposed criteria for the assessment of low frequency noise disturbance

Revision 1 December 2011 Contract no NANR45

NANR45: Criteria Revision 1, December 2011

Acoustics Research Centre, University of Salford

Document control

Version Issued Title Comments

NANR45 February 2005 Proposed criteria for the

assessment of low

frequency noise

disturbance

NANR45

revision 1

December 2011 Proposed criteria for the

assessment of low

frequency noise

disturbance

Corrections made to

Figures 35-37, Table 8

and text on page 58-63

ERRATA: revisions to NANR45 revision 1 issued December 2011

Section NANR45 Feb. 2005 NANR45 revision 1 December

2011

Figure 35 Distributions corrected

Table 8 Values corrected

Figure 36 L10-L90 values on x axis corrected

Figure 37 Plotted level instead of rate of

change of level

Page 58 This penalty does not go on

increasing as the L10-L90

increases, but „bottoms out‟ above

L10-L90 greater than about 6dB.

There is a transition region for

L10-L90 of between 4 and 6dB

where the penalty varies on a

sliding scale between 0 and 5dB

(as marked in dotted lines).

This penalty does not go on

increasing as the L10-L90 increases,

but „bottoms out‟ above L10-L90

greater than about 4 dB. There is a

transition region where the penalty

varies on a sliding scale between 0

and 5dB (as marked in dotted

lines).

Page 58 L10-L90<5: no penalty

L10-L90≥5: penalty of 5dB.

L10-L90<4: no penalty


Pages 59-

63

Changes in text for consistency

with the above


Acoustics Research Centre, University of Salford Page 1 of 114

CONTENTS SUMMARY ................................................................................................................... 3

INTRODUCTION ......................................................................................................... 4

Aim .................................................................................................................... 6

Review of existing criteria ................................................................................. 6

Sweden ................................................................................................... 6

Denmark ................................................................................................. 7

Netherlands ............................................................................................ 8

Germany ................................................................................................. 8

Poland .................................................................................................... 8

Comparison of national methods ........................................................... 9

Research Methodology .................................................................................... 12

FIELD STUDIES ......................................................................................................... 13

Details of the tests ............................................................................................ 13

Selection of case studies ...................................................................... 13

Measurement setup .............................................................................. 14

Equipment ............................................................................................ 14

Instructions to participants ................................................................... 14

Interviews ......................................................................................................... 15

Current and previous occupations ........................................................ 15

Current and previous address ............................................................... 16

Respondent‟s routine ........................................................................... 16

Health ................................................................................................... 17

Details of other people who hear the LFN ........................................... 19

Previous addresses ............................................................................... 19

Location of noise in and around the property ...................................... 19

Noise descriptions ................................................................................ 20

Sources of LFN .................................................................................... 20

Exposure to LFN .................................................................................. 20

Ambient noise level in home – expectation and control ...................... 21

Subjective reaction ............................................................................... 21

Noise avoidance ................................................................................... 21

General comments on interviews ......................................................... 22

Measurement results and analysis .................................................................... 22

Case 20 ................................................................................................. 23

Case 2 ................................................................................................... 28

General comments on field studies .................................................................. 34

LABORATORY TESTS ............................................................................................. 36

Objectives of the tests ...................................................................................... 36

Overall methodology for laboratory tests ........................................................ 36

Details of the tests ............................................................................................ 37

Selection of sounds .............................................................................. 37

Choice of subjects ................................................................................ 40

Length of the tests ................................................................................ 41

Listening room test setup ..................................................................... 42

Calibration of the listening room ......................................................... 43

Audiometric tests ................................................................................. 43

Test procedure ...................................................................................... 44

Laboratory test results ...................................................................................... 44

Low frequency hearing thresholds ....................................................... 44



Threshold of acceptability for pure tones ............................................ 46

Threshold of acceptability for real sounds ........................................... 49

Threshold of acceptability for „beating‟ tones ..................................... 53

Evaluation of fluctuations ................................................................................ 55

Fluctuation strength ............................................................................. 55

Standard deviation of sound pressure level ......................................... 55

„Prominence‟ ........................................................................................ 57

Conclusions from laboratory tests ................................................................... 59

CONCLUDING REMARKS ....................................................................................... 60

Proposed criteria and procedure for assessing low frequency noise ................ 63

REFERENCES ............................................................................................................ 64

ACKNOWLEDGEMENTS ......................................................................................... 65

APPENDIX: SUMMARY OF RESULTS FROM FIELD STUDIES ........................ 66

Case 5 ........................................................................................................................... 66

Case 6 ........................................................................................................................... 70

Case 7 ........................................................................................................................... 74

Case 8 ........................................................................................................................... 78

Case 13 ......................................................................................................................... 82

Case 16 ......................................................................................................................... 86

Case 18 ......................................................................................................................... 93

Case 19 ......................................................................................................................... 98

Case 19a ..................................................................................................................... 102

Control Case 1 ............................................................................................................ 106

Control Case 3 ............................................................................................................ 110



SUMMARY The aim of this study is to recommend a method for assessing low frequency noise

(LFN), suitable for use by Environmental Health Officers (EHOs) in the UK. A

general introduction to LFN is given, in which it is argued that a method of

assessment is needed both from the sufferer‟s point of view, because there is currently

not much to protect them against LFN, and from the Environmental Health Officer‟s

point of view, where guidance is needed in determining whether a nuisance exists.

Criteria already in use in Germany, Sweden, Denmark, the Netherlands and Poland

were reviewed and compared. Experience from these countries in applying the criteria

was also reviewed, and was found to be generally positive.

A complementary set of field and laboratory studies was conducted in order to

establish the best form for an assessment method. In the field studies, eleven cases of

reported LFN were investigated, as well as five control cases where no complaints

about LFN had been received. Analysis of recordings made over three to five days at

each location distinguished three groupings: positively identified LFN, unidentified,

and marginal. Three cases were positively identified, meaning that the various

national criteria were exceeded and there was correlation between the resident‟s

logged comments and the LFN level. Five cases were unidentified: the criteria were

generally not exceeded, (except perhaps by traffic noise), and there was a lack of

correlation between comments and noise levels. Three cases were marginal in that the

LFN was marginal with respect to the criteria and did not correlate with comments. It

was concluded that the criteria were successful at distinguishing cases where an

engineering solution could be applied from those where no such solution could be

found.

In the laboratory tests, a set of „thresholds of acceptability‟ were established by asking

18 subjects to set the level of various low frequency sounds to a just-acceptable level

for imagined day and night situations. The sounds presented consisted of a set of tones

across the low frequency range, „real‟ LFN extracted from field test recordings, and

synthesised beating tones with varying degrees of fluctuation. LFN sufferers were

found to be the least sensitive group in absolute terms, contrary to the common image

of ultra-sensitive individuals. In relative terms however, they were the most sensitive

group in that they set acceptability thresholds closer to their threshold of hearing.

From the existing national reference curves, the Swedish curve showed the best

agreement with the results. It was also demonstrated that fluctuating sounds are less

acceptable than steady sounds for the same average acoustic energy and should be

penalised. Furthermore, it was shown that 5dB is an appropriate penalty almost

irrespective of the degree of fluctuation above a limiting value.

A method for assessing LFN suitable for use by EHOs is proposed. This consists of a

reference curve based on 5dB below the ISO 226 (2003) average threshold of

audibility for steady sounds, plus a means to establish whether a 5dB relaxation for

steady sounds should be applied. It is expected that this will benefit EHOs by helping

to identify cases where they are able to improve the situation by enforcing noise

control measures. It is also expected that in a significant proportion of LFN cases it

will not be possible to identify a „hardware‟ solution. Consequently, it is suggested

that further research be conducted into alternative solutions.



INTRODUCTION In this section is given a brief introduction to the problem of low frequency noise

(LFN). There is no need for more than a brief introduction since a comprehensive

review was recently completed by Dr Geoff Leventhall et al. as part of a Defra funded

project [Le03].

Low frequency noise is now a recognised problem in many countries in the world.

Experience has been accumulating over more than 30 years, as a result of which a

picture has built up of typical situations where disturbance occurs. A relatively small

number of people are affected, but those who are tend to suffer severe distress. In

most situations, only a single sufferer, or perhaps a couple living in the same property

are affected. Occasionally, a cluster of complaints arises in a particular area, although

typically only a small proportion of people living in the area will report problems.

Although this picture is now becoming reasonably well-formed, this does not mean

that the causes of such suffering is fully understood, and many cases still go

unexplained.

There is a highly consistent vocabulary used by complainants, who may describe for

example “pressure on the ears” or a sound like “a diesel engine idling in the distance”.

Complainants frequently describe a sound that is intense, even deafening to them,

while many visitors to their home may be unable to hear it. It is common that they

also report a sensory perception of vibration (see for example [Mo02]) not perceived

by others. Visitors typically include local Environmental Health Officers (EHOs, who

have a statutory duty to prevent noise nuisance), water, gas and electric utilities, and

others. This discrepancy between how the sufferer and other people perceive the

sound can be one of the most baffling aspects of low frequency noise, and can leave

the sufferer increasingly isolated and confused. Many times a complete breakdown in

communications has occurred, the EHO being convinced that the complainant was

suffering from tinnitus1, and the sufferer equally convinced that the EHO was in

collusion with whoever was thought to be causing the noise.

Nowadays, thanks to an increasing number of documented cases, there is more

recognition of such cases, and a better understanding of how such situations could

occur. Fewer sufferers are misdiagnosed as having tinnitus and the knowledge that

other people around the world are involved in similar situations can be reassuring to

both the sufferer and the EHO.

How is it that one person could describe a sound as loud while another cannot even

hear the same sound? One possible explanation is based on the way the human

hearing system operates at low frequency. The perceived loudness of low frequency

sounds increases very rapidly with increasing acoustic energy. Therefore, low

frequency sounds only just above the threshold of hearing2 can be perceived as loud,

even uncomfortably loud. Added to this is the fact that individual hearing thresholds

1 Tinnitus: ringing in the ears which can occur when there is no external sound present

2 Threshold of hearing: the level of the lowest sound that can be heard. This varies with the pitch or

„frequency‟ of the sound, the human ear being less sensitive at low frequency than at mid and high

frequency.



vary, so that people with more sensitive hearing can hear sounds inaudible to others.

Putting these two facts together we may find a situation where a low frequency sound

is above one person‟s threshold, enough to sound relatively loud, whereas another

person with less sensitive hearing cannot hear it. This situation does not arise with

most other (not low frequency) sounds, because their perceived loudness increases

much more slowly with increased acoustic energy. In other words „normal‟ sounds

need to have very much more acoustic energy than the hearing threshold before they

become uncomfortably loud. The experience of low frequency sound can therefore be

„counterintuitive‟, i.e. it may contradict our more usual experience of sound.

This by no means explains all cases. However, an appreciation of the above subtleties

is extremely important because the counterintuitive nature of low frequency sound

makes it difficult to base accurate judgements on personal experience. Therefore, the

more widely understood these ideas are the better.

An additional factor is that „sensitisation‟ to low frequency sound often occurs over

time, leaving the sufferer more aware of the sound and unable to shut it out or get

used to it. Instead, the sound may grow in importance until it can become all-

consuming. It is not fully understood why, but this effect tends to happen more with

low frequency sounds than other sounds. Therefore, a brief visit to a property affected

by low frequency noise does not always give an adequate impression of what it is like

to live with the sound, making evaluation even more difficult.

The preceding paragraphs give the idea that the effect of low frequency sound is

different in several ways to other types of sound. Although the number of cases is

only a small fraction of total noise complaints, the distress suffered can be

disproportionately severe. It is not uncommon for sufferers to sleep in a garden shed,

a garage, a car or a hotel to escape the noise. Some move if they can. Complainants

frequently describe loss of sleep and in the worst cases can contemplate suicide.

From the EHO‟s point of view, low frequency noise problems can take up a

disproportionate amount of time and resources. They are notoriously difficult to tackle

even for specialists with long experience of this type of problem and with good

equipment. Added to this, not all acoustic instruments used by local authorities are

suitable for low frequency noise evaluation since the vast majority of noise cases do

not need a low frequency capability. Unfortunately many cases end up with a

breakdown in communications between the EHO, who often goes to a great deal of

trouble on the sufferer‟s behalf but is unable to detect the source of the problem, and

the sufferer, who is convinced the EHO is doing nothing.

The success rate of solved cases is not high, and unsolved cases tend to remain „open‟

for a long period, often several years. This is unsatisfactory both for the EHO, to

whom such cases can become an open-ended burden on resources, and the sufferer,

who may be left in a state of expectation but with no real prospect of a solution.

Sadly, a relatively high proportion of such cases end up with an investigation by the

Ombudsman, which puts both sides under a great deal of stress but rarely leads to a

satisfactory solution.

Even when the local authority is convinced there is a statutory nuisance and is able to

locate the source (they can only serve a notice if they know who is causing the



problem) they are often reluctant to take the case to court. This is because there is a

lack of authoritative guidance to support their case, and without such support the case

is not at all certain to be successful. For this reason, local authorities have been known

to find some other, less controversial grounds for serving a notice rather than base

their case on the uncertain legal territory of low frequency noise. Therefore, the

current situation is that local authorities need considerable resolve and, one might

even say, some courage to consider a prosecution for low frequency noise.

From the above, it is clear that some authoritative guidance would benefit both the

sufferer and the investigating EHO. This is needed both to help the EHO to identify

genuine problems more quickly and to support enforcement when a nuisance exists.

Aim

Hence, we come to the aim of this report, which is to recommend a method for

assessing low frequency noise, suitable for use by Environmental Health Officers in

the UK.

It is most important that any such method is fair. If noise limits are set too high then a

proportion of the population is not protected, and if set too low then an unfair burden

is placed on industry (in cases where sources of low frequency noise can be identified

they are usually industrial). In view of the well-known technical difficulties in

assessing and evaluating low frequency noise, as well as the complexity of human

reaction to sounds, this is not a simple task and needs to be done with considerable

care.

The guidance is intended to cover low frequency noise from industrial, commercial

and domestic sources, in particular rotating machinery but also including for example

combustion noise and turbulence. Music noise is not included.

Review of existing criteria

In pursuing the aim of the project, we can take into account the growing body of

experience about low frequency noise. Of particular relevance is the experience from

other countries where low frequency noise criteria have been adopted. Therefore, the

relevant authorities in countries with existing criteria were followed up in order to

evaluate their experience. In this section is given a brief review of existing criteria,

plus a summary of the reported experience from countries using them. Full

discussions of the criteria are also given in references [Le03] and [Po03].

Sweden

The Swedish guidelines state that low frequency noise should be assessed by third

octave band measurements in the range 31.5-200Hz. The sound pressure levels given

in Table 1 and Figure 1 should not be exceeded in any third octave band.

A survey of local authorities in Sweden was carried out recently and it was

ascertained that 62% of local authorities found the method to be better or much better

than the previous method. Of the remainder, 35% said they did not know, and in most



of these cases this was because they did not have the equipment needed to follow the

procedures. Only one local authority (3% of the sample) said they thought the method

was worse than the previous method. The positive response was received despite the

fact that the method is more difficult to apply than the previous method, more time

consuming and requires greater competence and equipment. This indicates that EHOs

(at least in Sweden) see the extra effort involved in the assessment as a worthwhile

investment.

Denmark

In the Danish method sound measurements are taken at several positions throughout

the property in the low frequency third octave bands (see [Po03] or [Le03] for a

description). Only the low frequency bands from 10-160Hz are included which, in a

normal A weighted measurement, tend to be de-emphasised and dominated by higher

frequencies. The measured third octave band values are then A-weighted and summed

together to give a low frequency, A-weighted level, LFpAL , . This value is then

compared with limit values given below. There is no reference curve as such in the

Danish method, but it can be compared with other criteria by assuming that the sound

is all concentrated in one third octave band. This gives the values given in Figure

1and Table 1. In practice this extreme situation does not arise, so although this

assumption allows us to compare with other criteria it gives values that are artificially

high. Another feature of the Danish method that differs from the approach in other

countries is that it specifies a 5dB penalty for impulsive sounds.

Maximum acceptable levels for LFpAL , are specified for certain areas:

Dwellings evening/night (18h-07h) 20dB

Dwellings day (07h-18h) 25dB

Offices/ teaching rooms 30dB

Other work rooms 35dB

Experience from the Danish Environmental Protection Agency indicates that the

limits are rarely exceeded, but when they are, the Local Authority is usually able to

locate the premises responsible for causing the noise and serve a notice or otherwise

regulate the situation. These guidelines are not considered to be entirely satisfactory

as they are relatively complicated for the EHOs to apply. Although there is little

quantitative information on whether the limits are set at the right level the general

feeling is that they are set are „close to OK‟.

A report by Sorensen [So01] suggested that the 20dB limit for LFpAL , was strict if the

noise was at the lower end of the range, i.e. 10-30Hz and unconservative for sounds at

the higher end, i.e. around 160Hz. This was for multi-tone spectra as are often found

with reciprocating engines in power plants, but is based on only two case studies and

is not intended to give a firm, general conclusion.

A separate criterion for infrasound also applies in Denmark, although cases of

infrasound are reportedly extremely rare. When the infrasound limits are exceeded

then low frequency limits are often also exceeded at the same time. The low

frequency limits are therefore seen as more important, whereas infrasound is not seen

as an important environmental concern in Denmark.



Netherlands

This method is intended to determine whether suspected LFN is audible or not, rather

than whether it should be classed as a nuisance [vB99a]. Audibility is based on

hearing thresholds for the 10% most sensitive people in an otologically unselected

population aged 50-60 years. These 10% thresholds are typically about 4-5dB lower

than the average threshold for otologically normal young adults (18-25 years) as given

in ISO226 [IS03].

Experience of the guidelines is generally positive. Consultants and EHOs are

generally aware of them and use them to assess potential problems. It is reported that

in not all cases where complaints occur is the threshold exceeded, and even in some of

these cases there is not a clear correlation between the reported disturbance and the

presence of the source. Arguably they are too low, because most complainants have a

higher hearing threshold than that given in the limits, but this has not been

systematically evaluated. Other investigators are more confident that the Dutch levels

are neither too low nor too high.

Germany

In the German method [DI97], a simple preliminary measurement is recommended in

order to determine whether the problem should be investigated further. If the

difference dBC-dBA is greater than 20dB then third octave band measurements

should be taken. The third octave readings are then compared with the values given in

Figure 1 and Table 1. (The values in Table 1 and Figure 1 are equal to the reference

curve up to 63Hz, but include corrections of +5dB at 80Hz, and +10dB at 100Hz.

These are the amounts by which the reference curve may be exceeded by tonal

sounds, so they have been added in the table to ease comparison). Different

procedures apply if the noise is tonal or not. The noise is said to be tonal if the level in

third octave band exceeds the levels in the two neighbouring bands by more than 5dB.

A tonal noise that exceeds the values in Figure 1and Table 1 at night time is

considered to be a nuisance. A 5dB increase in all bands is allowed for day time

exposure.

If the noise is not tonal then a day time limit of 35dB is imposed on the A weighted

equivalent level (10Hz-100Hz), where the A weighting is obtained by only using the

third octave bands for which the threshold is exceeded. The night time limit is 25dB.

The analysis required is therefore relatively involved for the non-tonal sounds.

Poland

The Polish method [Mi01] uses a reference curve defined over the range 10-250Hz,

and denoted LA10 because at each frequency it has the value equal to a pure tone of

10dBA. It is shown in Table 1and Figure 1 and is the lowest of all the curves at 50Hz

and below, and is also below the hearing thresholds as defined in ISO226 [IS03]. Low

frequency noise is considered annoying when the sound pressure levels exceed the

reference curve and simultaneously exceed the background noise level by more than

10dB for tonal noise and 6dB for broadband noise.



Although it may seem excessive to set a reference curve below the threshold of

hearing, this was justified [Mi01] on the basis that subjects in a laboratory test could

hear combinations of tones at lower levels than they could pure tones. Since the

published thresholds are based on pure tones it was argued that LFN consisting of

multi-tones (which it commonly does) may be audible at levels below the published

threshold. However, this argument has not been incorporated explicitly in other

national guidelines.

Comparison of national methods

The various reference curves are shown in Table 1 and Figure 1. Note that the Danish

curve, is applied in a different way to the others, so it is not strictly correct to compare

on a frequency by frequency basis. The impression that it is higher than other curves

is therefore slightly misleading from the figure. Also, the Polish curve, which appears

lower than others, is to be applied with an extra condition on background noise, and

for this reason is closer to the other curves than appears. The Netherlands curve is

intended only to predict audibility rather than acceptability, and so is lower.

Therefore, the national reference curves actually show more agreement than appears

from Figure 1and Table 1.

There are differences in the frequency range covered by the various curves. The

lowest frequency is set at 10Hz in the Danish, Polish and German methods, although

the German method includes an optional extension down to 8Hz. The Dutch and

Swedish methods start at 20Hz and 31.5Hz respectively. Thus, there is not complete

agreement on the lowest frequency that should be included. At the high end of the

range, the German and Dutch methods stop at 100Hz, whilst the Danish, Swedish and

Polish methods end at 160, 200 and 250Hz respectively. However, all the reference

curves rise away from the threshold of hearing above 63Hz, so that the bands at the

top of the range are significantly de-emphasised. Therefore, in effect there is more

agreement than appears in that all methods give most importance to frequencies up to

100Hz.

In terms of the levels of the curves, there is almost complete agreement that sounds

should be inaudible between 31 and 50Hz (a range where many reported problems

occur). However, there are some differences in the interpretation of audible.

A review of the various national criteria was recently carried out by Poulsen [Po03].

He played various sounds to subjects in a laboratory and carried out an analysis to

find which of the methods was the best predictor of their adverse reactions. He found

that the best correlation was obtained for the Danish method, closely followed by the

Swedish method. The Danish method was superior only when it came to evaluating

impulsive noise such as from music. In this respect the Danish method includes a

penalty of 5dB for impulsive sounds but there was no objective indication of when

this should be applied. Since music noise is not included in this investigation the

Swedish method can therefore be considered to be as good. The German non-tonal

method also worked well.

In terms of ease of use, the methods requiring third octave band values to be summed

(Danish, and German broad band method) are more difficult to apply. The summing

operation itself is fairly straightforward and could be handled by the majority of



EHOs. However, the disadvantage is that this cannot easily be done in real time on

site. This means the EHO evaluating the noise will not be able to get a feel for the

problem by making a quick assessment on site. This is actually quite an important

consideration since LFN problems are relatively uncommon, and EHOs often lack

confidence in their assessments due to lack of experience. For these reasons, methods

that specify a maximum third octave band value are preferred, because the

investigator can see what is going on more quickly.

The Polish method is the only one to require an assessment of the background noise.

This makes good scientific sense since it is often the case that background noise, e.g.

from traffic, is dominant in the higher frequency bands (between 100 and 200Hz).

Therefore, requiring the offending noise to be above background noise is a sensible

way to avoid a false classification of a normal background noise as a problem low

frequency. The drawback is that in practical situations it will rarely be possible to

measure the background noise. This is because for a background noise measurement it

is necessary to switch off the source being investigated. However, in a high proportion

of LFN cases the source is not known. Indeed, one of the purposes of an assessment

method would be to help to identify the source by narrowing down the problem to a

particular frequency band. For these reasons the Polish approach of including

background noise, although logically sound, is unlikely to be as useful to EHOs as

other simpler methods.

Compared with other environmental noise standards it may initially seem too stringent

to require levels of low frequency noise to be reduced to around the threshold of

hearing. However, there is a growing experience that such low limits are needed to

provide adequate protection from LFN. This is because of the strong reactions and the

apparent difficulty in habituating to LFN. The fact that all national criteria (with the

possible exception of the Danish one) are set at or below average hearing thresholds

gives strong support to this idea. It should be remembered that the standard threshold

is the median i.e., 50% of people are less sensitive and 50% are more sensitive. The

standard deviation of the experimental subjects tends to be around 6dB. Thus, about

16% of people have a threshold which is 6dB or more lower than the median, which

includes about 2% who have a threshold which is 12dB or more lower.

To summarise, on the basis of experience, the Swedish, Danish and Dutch (audibility)

methods appear to have all been positively received. On the basis of a laboratory

investigation the Danish, Swedish and German (non-tonal) methods were the best

predictors of annoyance. From the point of view of ease of use the Swedish, Dutch

and German (tonal) methods are most advantageous.



Hz Germany Denmark Sweden Poland Netherlands ISO threshold

8 103

10 95 90.4 80.4

12.5 87 83.4 73.4

16 79 76.7 66.7

20 71 70.5 60.5 74 78.5

25 63 64.7 54.7 64 68.7

31.5 55.5 59.4 56 49.3 55 59.5

40 48 54.6 49 44.6 46 51.1

50 40.5 50.2 43 40.2 39 44

63 33.5 46.2 41.5 36.2 33 37.5

80 33 42.5 40 32.5 27 31.5

100 33.5 39.1 38 29.1 22 26.5

125 36.1 36 26.1 22.1

160 33.4 34 23.4 17.9

200 32 20.9 14.4

250 18.6 11.4

Table 1: Reference curves used in the various national criteria, together with ISO threshold

Reference curves used in the various national criteria compared with ISO

threshold

0

20

40

60

80

100

120

8 10 12.5 16 20 25 31.5 40 50 63 80 100 125 160 200 250

Third octave band centre frequency, Hz

Level d

iffe

ren

ce, d

B

Germany Denmark Sweden

Poland Netherlands ISO threshold

Figure 1



Research Methodology

From the above discussion it appears that a useful criterion can be based on a

reference curve giving acceptable levels of sound at third octave band frequencies in

the low frequency range. Some objective method to assess the effect of fluctuations,

which appear to increase adverse reaction, would also be advantageous.

The form of the reference curve has been discussed above. Most existing curves are

based on thresholds of audibility, which have been established for many subjects over

many years, and provides us with the most comprehensive and reliable data about

hearing in the low frequency range. Regarding fluctuations, there is much less data

available. It is not possible to determine the effect of fluctuations through field

studies; for one thing it would not be practicable to survey enough cases, and for

another, there is too much variation between field studies, including the personal

situation of the subjects, the length of exposure and the character of the sound. To

establish the effect of fluctuations we need to measure the reactions of several people

to the same sound, and this can best be done by setting up tests in the laboratory.

There are limitations in laboratory testing of low frequency noise. In particular, the

disturbance in the field often includes an element of „sensitisation‟ to exposure over

extended time, and this factor cannot be reproduced in the laboratory. Nevertheless,

the annoyance of a sound can be judged by most subjects after a few minutes

exposure [Le03], so despite this limitation, laboratory testing is a well-established

technique.

To summarise: in view of possible sensitisation over time, the only true test of a

criterion is in real situations, but the only way to establish the effect of fluctuations is

through laboratory testing. Therefore, the research methodology chosen is a

combination of field and laboratory testing which are described in the following two

sections.



FIELD STUDIES The overall aim of the field studies is to provide support in the way of field data for a

proposed criterion. Specifically this involved collecting data with which to test

proposed criteria, and to provide audio recordings for use in the laboratory tests.

Human reaction to sound is known to be dependent not just on the sound itself, but a

complex array of other factors like personal associations of the sound. Therefore, in

each field study the sound measurements were supported by questionnaires to

determine whether sociological or other factors might influence the results.

Details of the tests

Selection of case studies

Cases were solicited through Environmental Health Departments by circular letter,

and by specific approaches to Local Authorities known to have a problem in their

area. More than forty cases were evaluated. EHOs who offered cases were approached

by phone and asked for a detailed description of the case. In some cases it was also

appropriate to approach the complainant at this stage. A few cases also came in by

word of mouth directly from sufferers.

Cases where several complaints occurred in a cluster were selected in preference over

those where a single complainant lived alone. This was because it is easier to justify

the complaints as reasonable if there are more than one. Also, it is well-known that

„mystery‟ cases often arise where no problem can be identified from recordings, and it

was thought that selecting clusters would help to avoid such cases.

Cases where there was a long history to the problem, particularly if there had been

modifications to a noise source during that time, were generally avoided. This is

because such cases can become overlaid with complications that make it more

difficult to know if the responses are purely due to the noise. For example, a number

of cases were received in which a low frequency noise source had been identified and

noise control work had been carried out to the satisfaction of most residents, but

where a smaller number had continued to complain afterwards. One possible cause of

this is that the complainants had become sensitised whilst the noise was present.

Whilst such sensitisation is a genuine part of low frequency noise cases, it becomes

more difficult to classify the response as typical and so stronger conclusions could be

obtained by excluding such cases.

Cases where the complainant was felt to be reasonably objective and perceptive in

their judgement of the sound were selected where possible.

EHOs and sufferers alike were generally keen to participate. Both groups were told

that we were not intending to solve their particular problem, but rather to contribute to

improved methods of evaluation in general. We adopted a policy that data collected

would not be released to either party, since this could have caused political



complications. Whilst all were generally anxious to solve their problem (which in

most cases had defied resolution), they were generally happy to participate on the

grounds that the results might help others in the future. Participants, both EHOs and

sufferers, were generally extremely co-operative and helpful.

Measurement setup

Although the majority of environmental noise standards specify that sound

measurements should be conducted outside, it is now generally agreed that low

frequency noise can only meaningfully be evaluated inside. All national standards

specify indoor measurements. Therefore, all measurements were carried out inside

complainants‟ homes.

A single microphone was positioned at a point in the room where the sufferer

indicated the sound was present. In most cases an unoccupied room was used. In two

cases an unoccupied bedroom was not available, so an occupied room was used,

although this was avoided if at all possible. In order to minimise data storage,

recordings were taken only when the sufferer said the noise was at its worst. In all

cases the noise, although usually present during the day, was reported worst at night.

Therefore, recordings were usually made between 21h00 and 09h00 when interference

from other sources is reduced. In some cases, at the request of the resident, recordings

were also made during the day. However, the most valuable recordings were all from

the night time period due to minimum interference from other sources. The equipment

was left to monitor unmanned for between 3 and 5 days.

Equipment

Measurements were taken using 01dB Symphonie systems. The signal from the single

microphone was simultaneously captured on two data channels which enabled all the

required parameters (dBA, dBC, third octave band levels and audio) to be monitored

simultaneously. The microphone and measurement chain were calibrated down to 1Hz

against a traceable standard in the UKAS accredited Calibration Laboratory at Salford

University immediately prior to the tests. In each location audio recording plus a wide

range of indicators, including one third octave band levels down to 1Hz were taken.

Data was streamed directly to hard disk and subsequently downloaded to DVD disks

for archiving.

Instructions to participants

Subjects were asked to complete a log sheet (see Figure 2) giving comments on how

they perceived the sound at particular times.



Please complete the following log sheet filling in the date and time when you hear the

noise being studied, rating the noise along the scale from „Not at all disturbing‟ to

„Intolerable‟ and adding any other comment you feel may be important. It is useful to

us to have information about when you are not disturbed as well as when you are.

Please use as many sheets as necessary.

Date and Time Rating (please place a tick along the line) Comments

Not at all

disturbing Intolerable |----|-----|-----|-----|-----|-----|-----|-----|

Figure 2: Log sheet given to subjects in the field studies

The terms „disturbing‟ and „intolerable‟ were used deliberately. Most studies of

environmental noise use the term „annoyance‟ to judge the severity of the response.

However, the descriptions and vocabulary used by low frequency noise sufferers does

not generally suggest that „annoyance‟ is their main concern about the LFN. The term

„disturbing‟ is considered to represent the response of a typical sufferer more

accurately and so was used on the log sheets. Again, the term „intolerable‟ was used

deliberately to help identify periods of extreme, unacceptable exposure, as perceived

by the sufferer, from the recordings.

Interviews

In addition to making physical recordings of the sounds within complainants‟

residences it was necessary to obtain a significant amount of personal data about the

individuals themselves. This was important in order to obtain an overview of the

background to the LFN complaint that might have a bearing on the responses. Details

were collected about each individual‟s residential and occupational histories, their

general health, details of the noise they are exposed to, suspected sources of the noise,

effects of the noise on themselves and their health, and any measures they have taken

to cope with or avoid the noise.

Using a comprehensive one-to-one structured interview schedule we obtained detailed

personal information from 12 LFN sufferers. All 12 sufferers answered all questions

without hesitation and were forthcoming and open when answering questions relating

to their general and mental health and when providing detailed information about their

noise problem. This section reports on the specific questions asked and the

information obtained.

Current and previous occupations

Details of current and previous occupations were gathered in order to determine

whether people had a work-related exposure to LFN. This could be relevant if

theories of sensitisation to LFN are verified. Additionally, it was important to



establish whether there was any employment-related connection to the main suspected

source of the LFN.

Only one complainant reported any previous work-related exposure to LFN and had

worked in the industrial works section at British Rail for 28 years since an

apprenticeship over 40 years ago. This had entailed exposure to heavy drop forges,

pneumatic air guns and general industrial machinery noise. This complainant made

some earmuffs as nothing was provided at the time. No other complainants reported

working in environments that could be considered likely to expose them to LFN.

None of the complainants named a previous employer as the likely suspect of the LFN

to which they are exposed.

Current and previous address

Details of current and previous addresses were gathered in order to determine how

long complainants had resided at their current address and whether their residence

predated exposure to LFN. Additionally, it enabled clarification to be sought on

health related matters in later questions.

Two complainants stated that the LFN was detectable when they first moved into their

house but all others had many years of no exposure prior to the onset of the problem.

This ranged from 4 ½ years to 38 years.

Respondent’s routine

Details of daily routines were collected in order to determine when the house was

busiest and quietest. Given that recording equipment was placed in the home for up

to 7 days it was necessary to determine when quality recordings might be obtained.

Knowing bedtimes and waking times enabled the recording devices to be set to record

at appropriate times.

Determining what woke the complainant during the night was important to determine

the amount of sleep deprivation associated with exposure to LFN. In some cases it

was not the noise that awoke the complainant but subsequent awareness of it

prevented them returning to sleep. Asking what the complainant did when wakened

in the night was to determine whether any additional coping strategies were employed

at that time. Later in the report we discuss the main coping strategies used by

complainants not just those employed in the night-time.

All bar one respondent wakes during the night and 75% of all respondents claim it is

the noise that wakes them. Over half of complainants get up and walk about when

they wake up and a quarter look around the house or out the window to try to

determine the source of the noise that has disturbed them. Other activities include

putting on the television, using the bathroom, making a drink, taking a sleeping tablet

or putting in earplugs. A third of all respondents sometimes just lie in bed listening to

the noise without taking any other action, although they may take action on other

nights.



Health

Complainants were asked personal questions about their general health to help

determine what health related problems they suffered from. Initially, at the beginning

of the interview, complainants were asked to self-report symptoms they suffered from,

both related to and unrelated to the LFN problem, to give them the opportunity to list

what they considered the most significant health issues in their lives.

In order to determine whether complainants might have a hearing problem they were

asked whether they had ever had their hearing tested, how long ago this took place,

and the outcome. They were also asked whether they were satisfied with the

outcome. This was done in order to rule out hearing problems as a cause of the

problem. Not every complainant had had a recent hearing test done. Complainants

were also asked if they had ever suffered from tinnitus.

A quarter of complainants said they had known hearing problems. One had a 60%

hearing loss in one ear with the other ear normal. Another had age-related hearing

loss with a loss in the higher frequencies, and one had a blockage due to sinusitis

which produced a whistling in the ear.

Half of complainants had never had a hearing test and only 2 had had one within the

previous year.

All knew what tinnitus was when asked whether they had suffered from it. All bar

one said they had never suffered from it and one said they were not sure as they did

sometimes get a whistling in their ear. This was attributed to sinusitis.

Finally a list of other symptoms was read out to the complainant and they were asked

to state whether they suffered from any of them. It was made clear that they should

say whether they suffered from the symptom whether or not they attributed it to

exposure to LFN. The list of symptoms was obtained from the published literature on

LFN exposure with particular reference to Leventhall (2003) [Le03].

Respondents who practiced successful coping strategies were asked to report health

problems at the time when the noise was at its worst. Our intention was to obtain a

list of symptoms experienced by sufferers although we do not have sufficient data or

expertise to determine aetiology. Further research in this area is required.

The table below shows the results of this line of inquiry listed in the order in which

the questions were asked. 92% of complainants suffer from sleep disturbance, 83%

suffer from stress and 67% have difficult falling asleep. 42% suffer from insomnia

and 33% from depression. 33% suffer from palpitations although none claim to have

heart ailments. 58% suffer from headaches and 25% from migraines. 42% have high

blood pressure. Perhaps most seriously, 17% have felt suicidal.

Health

Number of respondents

Percentage of respondents

Tinnitus *1 8%

Stress 10 83%

Loss of concentration 6 50%

Sleep disturbance 11 92%



Difficulty falling asleep 8 67%

Frequent irritation 5 42%

Nausea 1 8%

Nervousness 6 50%

Insomnia 5 42%

Chronic fatigue 2 17%

Anxiety 8 67%

Frustration 9 75%

Depression 4 33%

Indecision 1 8%

Tiredness 7 58%

Exhaustion 3 25%

Dizziness 2 17%

Sinusitis 3 25%

Glaucoma 0 0%

Pressure or pain in ear or body 7 58%

Body vibration or pain 6 50%

Palpitations 4 33%

Heart ailments 0 0%

Frequent ear vibration 2 17%

Eye ball or other pressure 3 25%

Pains in neck 5 42%

Backache 2 17%

Migraine 3 25%

Headaches 7 58%

Subdued sensation 1 8%

Shortness of breath 2 17%

Abdominal symptoms 3 25%

Shallow breathing 3 25%

Chest trembling 1 8%

Hypertension 5 42%

Stitch 1 8%

Difficulty reading 4 33%

Difficulty watching tv 4 33%

Difficult listening to radio 1 8%

Head injury 2 17%

Dental disease / surgery 5 42%

Eye surgery 0 0%

Suicidal 2 17%

* respondent attributed whistling in ear to sinusitis rather than tinnitus

Table 2: Numbers of respondents reporting various symptoms

The list of health questions included details of surgery and dental treatment undergone

in order to identify any possibility that symptoms may be related to head injuries or

dental surgery. 42% of complainants mentioned some form of dental surgery the

most invasive of which was root canal treatment received by one individual. The

other reports refer to tooth extraction, dentures and crowns. Two individuals reported

head injuries and these can both be discounted as sources of noise complaints. One

head injury was whiplash from an accident a year previously (this respondent has

suffered from LFN for 14 years), the other was from collapsing two years previously

due to an incorrectly prescribed dosage of medication which resulted in requiring

stitches across the head (this respondent has suffered from LFN for 5 years).



Details of other people who hear the LFN

Sufferers were asked about anyone else who heard the LFN to which they were

exposed as previous research has shown that LFN may be detectable by some people

and undetectable by others. All of our sufferers reported that other people had heard

their noise but that not everyone who came to their residence was able to hear it.

Overall a wide mixture of „others‟ could hear the noises including family, neighbours,

friends and other visitors to the house. Many also reported people saying that they

thought they could live with the noise – i.e. it didn‟t bother them as much as it did the

complainant.

Further to this we asked whether the noise annoyed every person who heard the noise

equally. This was in order to determine whether people had their own explanations or

theories about why they were bothered but others were not. Responses to this

included observations about other people being too busy to be bothered by it, having

more going on in their homes (family, children, loud music etc), and worrying about

the value of their home if they made a complaint and then couldn‟t sell their property.

Other people wondered whether they were more sensitive to LFN than others or

whether they were simply able to hear sounds at lower frequencies than other people.

Previous addresses

Sufferers were asked whether they had experienced similar LFN problems at their

previous address (even if this was a very long time ago) or at any time in the past

prior to the onset of the complaint in question. Our intention was to find out whether

they had been bothered by LFN when they were younger, although a negative

response to this question does not signify that younger people do not detect LFN.

Two complainants stated that they had had similar noise problems at previous

addresses. One of these was related to traffic noise, which caused the windows to

vibrate causing a hum. This was rectified with secondary glazing. The other was

attributed to living amongst factories and the complainant expected to hear noise at

that location.

Additionally complainants were asked whether their sleeping patterns were the same

at their previous address in order to see whether their current pattern was a constant

throughout their life. This question elicited more discussion of changes in lifestyles

and work patterns than answers related to LFN.

Location of noise in and around the property

Details of where in the property the sufferer heard the noise were gathered in order to

determine the best location to leave the recording equipment. Knowing whether the

noise was more detectable in particular positions in rooms enabled a more precise

location to be found, thus allowing the best obtainable recording.

Complainants were asked whether the noise was better or worse with the window

open as the literature suggests that LFN is exacerbated in enclosed rooms due to

windows and walls filtering out higher frequency sounds. It is sometimes experienced



that opening the window ameliorates the LFN problem, although only 25% of

respondents reported this. 50% of respondents said it made no difference whether the

window was open or closed.

Noise descriptions

Complainants were asked to describe in their own words the LFN to which they were

exposed. Subsequent to this they were read a list of further descriptions, from the

literature on LFN, to see if any of them also matched their noise. In an effort not to

put words into the complainants‟ mouths they were asked for their own descriptions

first.

In addition to the descriptions commonly used by complainants to describe LFN, as

reported in the literature, other descriptions used by our respondents included: „like a

car ticking over‟; „a distant hum‟; „like a refrigerator building up again after the door

has been opened and closed‟; „like a central heating boiler‟; „a whine like a jet engine

or turbine‟; „a whistle‟; „a short beat and a long beat‟; „like a lorry with the engine

going‟; „like a meter winding down‟; „like a spin dryer‟; „like being in a microwave‟;

„like a kettle warming up‟; „like aircraft high overhead‟; „a deep roar‟; „like a

compressor unloading‟; „like emerging from a tunnel‟; „like fishing boats going to sea

at night‟; „like air roaring up a chimney‟.

Sources of LFN

Complainants were asked if they knew the source of the noise. While a third of cases

said they did know the source they were only able to narrow it down to a site (a

particular commercial or industrial premises) rather than to a specific process or piece

of equipment. Two thirds of cases did not know the source but had a variety of

theories, usually with a favoured suspect. Details of how the source or potential

source of the noise was identified were gathered and these included visiting the sites

in question or obtaining information about when new equipment was brought online

at the sites.

Complainants were asked formal questions about the history of their LFN problem.

While it was felt unnecessary to obtain the complete detailed history of the

relationship between the complainant, the EHO concerned and the suspected source of

the LFN enough information was sought to establish how long the LFN had been

present and what steps had been taken to identify and rectify it. These histories often

identified long running problems with considerable involvement of the EHO.

Sometimes ameliorating procedures were put in place which either rectified part, but

not all, of the problem, or seemed to remedy the problem only for it to start up again

in subsequent years.

Exposure to LFN

In order to obtain good recordings of the LFN complainants were asked what time of

the day the noise was worst. It was expected that this would be at night time when

background noises are reduced although this was not always the case. Often

complainants attributed bad times to particular periods in the cycle of the equipment

or process considered to be the most likely source. A number of complainants



mentioned how wonderful it was when the noise stopped – sometimes for a fortnight

over Christmas or summer (which they attributed to the plant in question closing

down for holidays) – only they were on edge all the time expecting it to start up again

at any time.

For the majority of respondents (83%) the LFN was continuous, i.e. it was always

there. The remaining two respondents said their noise was intermittent, with silent

periods in between.

Ambient noise level in home – expectation and control

Respondents were asked how they would describe the background noise level in their

home taking the LFN out of the equation. Given that other sounds may mask LFN we

wanted to ascertain the extent to which masking sounds were present. All described

their home as „Quiet‟ or „Very Quiet‟. This raises questions about expectation and

whether some people have higher expectations of intrusion of noise from external

sources. However, some sufferers stated that they didn‟t mind aircraft flying

overhead or the sound of the road outside their home because they knew it was

intermittent and other stated that it was knowing the source that was important to

them as it gave them a sense of control.

Subjective reaction

Complainants were asked to state in their own words how the noise made them feel.

Put this way the question allowed for a repetition of the health symptoms the

complainant suffered from or a further description of their emotional response to the

noise. Many complainants spoke of the frustration they experienced, their lack of

control, and the lack of help or success from agencies they expected to have power

over the situation.

Noise avoidance

Details of any measures that had been taken to try to avoid the noise were obtained.

The importance of this was to identify the extremes to which people have gone to

avoid the noise as well as to identify measures that worked and could therefore prove

useful strategies for other sufferers.

Half of the sufferers interviewed have tried earplugs and some still use these at night.

Others said that they exacerbated the problem or made no difference. Headsets or ear

defenders have been tried by a third of sufferers, sometimes in tandem with earplugs.

Some complainants said that using head phones when watching TV aided their

concentration which was otherwise diminished.

Three quarters of the complainants had tried sleeping in different rooms in their house

with varying degrees of success. For those who found that one façade of the house

was worse than another moving to a „back‟ room proved successful. However others

found no respite despite trying to sleep in the living room, the hallway, the kitchen,

the cellar and/or the balcony. Some attempted putting foam under the bed legs, with

no effect, while others slept with their head pointing towards the middle of the room

rather than against the wall, with some effect.



One respondent said they would go away to a holiday home they owned in order to

avoid the noise and would extend their vacation because they hated coming home.

Nearly half had considered selling their home, some had even put it on the market, but

all were concerned about their duty to tell potential buyers about the LFN problem.

Those that had tried to sell found it impossible once potential buyers were aware of

the LFN problem.

A quarter of the sufferers said they tried to concentrate on other things in order to

divert their attention from the LFN. Techniques that came under this category include

practicing yoga and other stress reduction techniques.

A quarter of sufferers regularly took prescribed sleeping tablets in order to sleep and

found these very successful. Others were not willing to take sleeping tablets.

Creating additional noise to mask the LFN was tried by some, again with varying

success. Using a „tinnitus machine‟ worked extremely well for one complainant who

had discovered this on the internet. Similarly an air purifier worked for another

complainant. Another had found playing „white noise‟ on the radio helped them

sleep. However, playing the radio or TV during the night had limited success

although more during the day.

All respondents were asked if they had any additional comments that they felt were

relevant to their problem but which had not been covered in the interview. No new

information was obtained in this way.

General comments on interviews

The results presented above indicate that all the complainants used in the study have

ongoing problems which they associate with low frequency noise, and which have a

fairly serious impact on their lives. None have a history of suffering from these

problems at previous residences, and none have had an employment or other

discernable relationship with the company or organisation suspected as the source of

the low frequency noise about which they complain. Furthermore, as far as can be

judged by an experienced interviewer, the complaints were genuine, and there was no

hint of ulterior motives, such as wanting to get rid of local industry.

Responses to the problem of exposure to low frequency noise ranged from an annoyed

interest to feeling suicidal. Coping strategies ranged from wearing earplugs through

sleeping in different rooms to attempting to sell the house. Not all respondents had

found a strategy that worked for them at the time of the interviews although we were

able to pass on information about how other sufferers coped.

Measurement results and analysis

A large amount of data was recorded for each case study. This was considered

necessary, since from experience, the equipment must typically be in the property for

several days to capture a period when the complainants report hearing a representative

„bad‟ noise. One of the problems of LFN analysis is how to make sense of such a



large amount of data. The details of the analysis varied from case to case, but the

usual steps were as follows:

a. Several periods were selected from the subject‟s log about the time they said

the noise was particularly bad (the period was chosen to encompass the time

given by the occupant, but to exclude events such as doors closing etc. as

detected by ear)

b. For each such period a sonogram was drawn to display the 1/3 octave

spectrum. This was examined to see whether any events could be identified

that correlated with the respondent log. The sonogram option may not be

available to most EHOs, but a third octave band spectrum could be used

instead.

c. From the third octave band plot, the single third octave band that exceeded the

audibility threshold by the highest margin was selected

d. A narrow band plot was also made to see if there were any obvious tonal

frequencies in this band

e. A plot of the sound level in this third octave band was then plotted against

time so as to show what, if anything, happened at the time identified as being

bad.

In all but two cases it was possible to identify suitable periods described by the

subject as particularly bad. In Case 8 the subject did not make a detailed log, asserting

simply that their noise was present all the time. In Case 6 there was some question as

to whether it was the subject themselves or a spouse who had compiled the logs. For

these cases we selected the worst case situation by a combination of looking at the

spectra and analysis „by ear‟ of the audio recordings.

In this section are presented two cases which are illustrative of the other cases. A

summary of the results of other cases is shown in Appendix 1.

Case 20

Figure Times identified by respondent selected for presentation

below Figure 5 03h00 “Bad throb, headache, felt sick” Figure 8 00h30 “Very bad rumble” Figure 8 08h15 “Very bad throb”

Location Rural

Source Distant industrial

Microphone

position

Corner of downstairs living room

The narrow band plot in Figure 3 is from the around the time indicated as bad by the

complainant, and shows a clear pronounced peak at about 36Hz. Sounds at this

frequency, if of high enough intensity to be audible, would be heard as a low

booming. However, the presence of this peak by itself is not enough to demonstrate a

problem, we need to compare it with the threshold of audibility and with national

criteria. This is done in the third octave band plot, Figure 4, taken at around the same



time, on which is also shown the ISO threshold of audibility. The 40Hz band is above

the threshold by more than 20dB, and is the most likely candidate to cause a problem.

Having clearly identified the 40Hz band as the likely source of the problem, a plot of

the variation in the level of this band with time is shown in Figure 5. The time given

by the complainant is marked, and clearly corresponds with a time when the level in

this band was raised. Also shown in this figure for comparison are the limits from the

Polish and Danish national criteria, which are respectively the lowest and highest

values of any of the national criteria for this band. Note that the Danish curve is not

strictly intended to be used as a reference curve in this way, and so the value plotted is

if anything on the high side. Even without taking this into account, the levels are

above the curve, and therefore above all the national criterion curves.

Figure 6 shows the dBA and the dBC levels plotted during the night. dBA is the usual

indicator for environmental noise, and filters out low frequencies. dBC does not filter

out low frequencies. The amount by which the dBC level exceeds the dBA level

therefore gives an approximate indication of the low frequency content in the sound.

The difference is up to about 30dB, so the preliminary check used for the German

standard would show the need for a more detailed investigation. Note that after 07h00

the A weighted level rises significantly whilst the dBC level remains about the same.

This is due to activity of the residents getting up and making some sound in the house.

Thus, the sound is predominantly low frequency, except when there is „people noise‟

in the house.

Figure 7 shows the third octave band spectrum at another time indicated by the

complainant as particularly bad. It is similar to Figure 4 with the 40Hz band about

15dB above the threshold of audibility. The time history in Figure 8 shows that levels

exceed the Danish limits (and therefore all other national criteria) in the 40Hz band

throughout the night, and are particularly high at times indicated.

Therefore, in this case there is a clear correlation between times when the noise

exceeded guidelines and when the complainant reported being particularly disturbed.



[ID=153] Average G1 1 Hz;(dB[2.000e-05 Pa], PSD) 36.47 68.1

0

10

20

30

40

50

60

70

80

20 40 60 80 100 120 140 160 180 200 220 240

Figure 3: FFT of 9m30s audio record starting 02h28m from measurement

Case20_040508_210000.cmg

1 [Average] Hz dB40 70.0

0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k 16 k

Figure 4: Mean 1/3 octave band spectrum 9m30s starting 03h00m from measurement

Case20_458_210000.cmg



Figure 5: Time history showing 40Hz 1/3 octave spectrum band from measurement

Case20_458_210000.cmg together with lower Polish 44.6dB and Danish 54.6dB limits

10

15

20

25

30

35

40

45

50

55

60

65

70

75

80

21h 22h 23h 00h 01h 02h 03h 04h 05h 06h 07h 08h 09h


Case20_458_210000.cmg together with dBA levels. The dBA level illustrates normal household

noise.




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k 16 k


Case20_459_210000.cmg


Case20_459_210000.cmg together with lower Polish 44.6dB and Danish 54.6dB limits



Case 2

Figure Times identified by respondent as ‘Intolerable’ selected for

presentation below

Figure 11 23h45 5 on scale. „Hum louder through the night‟

Figure 14 07h00 2 on scale „Hum as usual‟

Location Suburban

Source Suspected industrial.

Microphone

position

By closed doubled glazed window in upstairs front bedroom

Figure 9 shows the narrow band recording from Case 2 at a time indicated by the

complainant as a score of 5 on the log sheet scale of 8 (Figure 2). This was the highest

that they recorded throughout the tests. Figure 10 shows the third octave band plot

with the ISO threshold of hearing superimposed. The only notable feature on both

plots is a peak at 10Hz, but this is more than 40dB below the threshold of hearing as

published in the German standard (the ISO published values do not extend down to

10Hz). The 80Hz band might be just audible, and the 100Hz band exceeds the audible

threshold by about 6dB, so could be audible. However, none of the bands up to 80Hz

exceeds any of the national criteria. The 100Hz band level is at a similar level to the

Polish curve, but would not exceed it when background noise was taken into account

as is required in the Polish method. Thus, none of the national criteria were exceeded

for this time.

The dominant source in the 80and 100Hz bands was road traffic noise. It is fairly

common to find audible noise in these bands due to traffic. This could be determined

by ear, and from the profile of the sound levels during the night (Figure 11) which is

typical for traffic. Figure 11 also shows the time the comment relates to. It can be seen

that this occurs at a time when the noise levels in this band are falling. (The

occasional „spikes‟ on this plot are due to internal movement or occasional events in

the neighbourhood, and are not associated with any steady noise of LFN type). The

description of the noise as „hum louder through the night‟ does not correlate with the

noise levels in this band.

Thus, for this time:

none of the national criteria were exceeded

the description given by the complainant does not correlate with the observed

variation in noise levels in the only band likely to contain audible sound

the noise in this audible band was due to road traffic.

The narrow band spectrum for another time is shown in Figure 12. There is a small

peak at 75Hz, which also shows up in the third octave spectrum Figure 13 (the source

is not known but is thought to be internal). This band is an average of more than 5dB

above the two neighbouring bands and slightly exceeds the German night time limit of

33dB. The time history in Figure 14 shows a similar typical profile of traffic noise

(again, the spikes are single events, not LFN). Noise levels are rising at the time of the

comment, but the complainant gives a score of only 2 on the scale of 8. These



findings do not correlate with the above comments from the earlier time where the

German criterion was not exceeded and a higher disturbance score was given.

Several other times and comments were evaluated, but we were unable to find a

relationship between noise levels and the comments. The cause of complaints in this

case therefore remains a mystery. In terms of the aims and objectives of this project it

provides a clear example of a case that could not be solved by engineering noise

control. This is true firstly because the only noise that could be identified was road

traffic noise. Secondly, even if a source could be found, the lack of correlation

between the respondent‟s comments and the presence of any raised noise levels

suggest that reducing noise levels would not resolve the complaints.




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case2_433_104426.cmg

Average G1 1 [Average] Hz;(dB[2.000e-05 Pa], PWR) 100 30.0

0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case2_433_104426.cmg




Case2_433_104426.cmg together with lower Dutch (audibility) 22dB and Danish 39.1dB limits


0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 12: FFT of 9m30s audio record starting 07h00m on 01/03/04 from measurement

Case2_4229_210000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 13: Mean 1/3 octave band spectrum 9m30s starting 07h00m on 01/03/04 from

measurement Case2_4229_210000.cmg





Similar analyses were carried out from measurements at nine other residences. Some

of the figures used in the analysis are given in the Appendix. The overall findings are

summarised in Table 3.

Case Peak

1/3

octave

band

(Hz)

Respondent

suspected

source

Correlation of

respondent log

with suspected

source

Source

indicated

by analysis

Correlation of

log with

analysis

indicated

source

2 100 Industrial No None N/A

80


63


100


80

8 50 Industrial No log Industrial N/A

50

13 100 Do not know N/A None N/A

100

16 100 Do not know No Air traffic

and

domestic

equipment

Yes

18 50 Industrial No Domestic

equipment

and possibly

air traffic

Yes

19 100 Industrial Yes Industrial Yes

50

19a 63 Industrial Yes Industrial Yes

63

20 40 Industrial Yes Industrial Yes

Control

1

63 Traffic Yes Traffic Yes

40 Domestic

equipment

Yes Domestic

equipment

Yes

Control

2

50 Domestic

equipment

Yes Domestic

equipment

Yes

Control

3


Control

4

160 Industrial Yes Industrial Yes


Table 3: Summary of findings from case studies and control cases. The multiple entries refer to

different events studied.



A number of control cases were also examined using the same techniques. These were

residences where low frequencies would be expected in the spectrum, but where there

had been no reports of disturbance, for example city centre residences and houses

with direct line of sight to a busy motorway. The findings are also summarised in

Table 3. Of particular note from these results is that in Control Case 1 the criteria

would have been exceeded due to a domestic central heating pump in the dwelling,

although there was no complaint about LFN.

General comments on field studies

The case studies fall into three categories:

a. Positively identified LFN – in these cases the national criteria were exceeded

and respondent logs correlated with recorded sound from an external source of

LFN

b. Unidentified – in these cases the national criteria were not generally exceeded,

(except perhaps by traffic noise or sound from internal domestic equipment)

and respondent logs did not correlate with any source

c. Marginal – in these cases a source of LFN could be determined but was

borderline with respect to the criteria.

The case studies falling into these categories are identified in Table 4

Table 4: Categorisation of case studies

Positively identified Case 20

Case 19

Case 19a

Marginal Case 16

Case 18

Case 8

Unidentified Case 6

Case 13

Case 7

Case 5

Case 2

In positively identified cases an engineering solution could be put into place, and in

view of the correlation with respondent logs, would be likely to remove the source of

the problem. In unidentified cases engineering solutions would not be possible, firstly

because no source could be identified, and secondly, even if it could, the lack of

correlation with the complainant logs suggest that the problem would not be solved by

reducing sound levels. In two out of three of the marginal cases the suspected source

involved air traffic, which would be beyond the control of a local authority.

Therefore, it appears that the national criteria are generally successful in

distinguishing between cases where the EHO is likely to be able to bring about a

solution from those where they are not. However, as control cases show, there is not

always a complaint when the criteria are exceeded. These conclusions are significant



in terms of the aims of this study; they imply that the criteria can be useful indicators,

provided they are not applied in a rigid way.

Most of the „problem‟ and „marginal‟ sounds were in the 40 and 50Hz bands. In these

bands the national reference curves are in reasonably close agreement, so the same

conclusion would be arrived at irrespective of the criterion used.

It was noted that in all cases the background noise levels in the residences were

extremely low apart from the LFN, if present. This is typical and has been observed

by various researchers (see for example [vB99b]). Such low levels of natural masking

noise are thought by some to be a factor contributing to the disturbance of LFN.

It is also noticeable that there were no cases in which the noise was reported to be

present only during the day. This does not mean that the noise was absent during the

day, most respondents said they could hear it during the day but that it was worst at

night. However, in every case the noise was reported to be present at night. This

contrasts with common experience, where a random batch of complaints about

general industrial noise (not LFN) might be expected to include some complaints

about industry that does not operate at night but causes disturbance only in the

daytime. This observation does not contribute to the main aims of this report, but it is

mentioned as being relevant to help explain the phenomenon.

A certain amount of judgement is involved in identifying LFN. One of the most useful

aspects of the criterion curves is to help identify problem frequency bands (see also

[Ru02]). A useful technique, which is now becoming more widely available, was

found to be to take audio recordings along with sound level measurements. It is a

common problem that the investigating person is hampered by not being able to hear

the sound themselves. Audio recordings can be played back at a higher (audible) level

and are useful to distinguish between various noise sources. Combined with third

octave and narrow band spectra, together with the criterion curve improves the

chances of being able to identify sources.



LABORATORY TESTS

Objectives of the tests

Much previous work, (including most national guidelines) is based on the idea that the

acceptability or otherwise of a low frequency sound can be evaluated in relation to a

frequency-dependent reference curve. This well-established approach will be adopted

here. Such a curve can be called the „threshold of acceptability‟: sounds with a higher

intensity would be considered unacceptable, and those with a lower intensity

acceptable. (The idea is similar to the familiar „threshold of hearing‟ which indicates

the level at which sounds become just audible rather than acceptable.) The overall aim

of the laboratory tests is to establish a threshold of acceptability for day and night time

and for sounds of various characters.

It has already been mentioned that it is not possible to reproduce realistic field

conditions in a laboratory test. In particular, the length of exposure does not give an

adequate impression of what it is like to live with the sound. Therefore, the laboratory

tests should not be used to establish absolute levels for a reference curve. However,

absolute levels have been fairly well established in the various national criteria and by

reference to published hearing thresholds, so this is not needed. What is needed is to

establish an optimum shape for the curve since the various national guidelines differ

in this respect. This will be the first objective of the laboratory tests.

Clearly, as with thresholds of hearing, the threshold of acceptability would be

expected to vary from one person to the next. It might also vary between day and

night time, and could show variation depending on the character of the sound as well

as its intensity. In particular, the degree of fluctuation in a sound has previously been

identified as an important parameter affecting the acceptability [Po03] [Le03].

The objective of the laboratory tests is therefore to establish „thresholds of

acceptability‟ for sounds with varying degrees of fluctuation, for day and night

exposure.

Overall methodology for laboratory tests

There are two general approaches to testing in the laboratory: the method of limits and

the method of adjustment.

In the method of limits a number of fixed sounds is played to the subject, who is

asked to give each one a „score‟ to indicate how much it annoys them, how pleasant

they find it etc. In the method of adjustment, the level of the sound is adjusted until it

achieves a certain response from the subject, for example it is adjusted so that they

can just hear it (this is how hearing thresholds are tested). In both methods one is

looking to find a correlation between an objective quantity (as measured by the

acoustic instrumentation) and a subjective quantity (as indicated by the subjects).



The method of limits is the best-established method for measuring reactions to

environmental noise (see for example Poulsen and Mortensen [Po03]). One advantage

for this study is that we could argue it is closer to real cases in that sufferers of low

frequency noise have no control over the sound, (other than to move to another room

or building). A disadvantage is that, since we are interested in establishing the effect

of fluctuations on the threshold of acceptability, we need as many sounds as possible

to be around the threshold level. This may not work out if we are fixing the level of

the sounds for all subjects since each individual will have a different threshold. This

leaves the possibility that some tests might not give useful data.

The method of adjustment could be used by asking subjects to adjust the level of the

sound so that it is just acceptable for an assumed situation, like trying to get to sleep.

(see for example Inukai et al. [In00] who carried out a series of tests on Japanese

subjects). The effect of fluctuations, if any, on the threshold of acceptability can be

judged directly from their responses to sounds of different character.

Therefore, the method of adjustment is well suited to the objectives of the tests and

was adopted. There are further advantages in that the comparisons can be done more

quickly than with the method of limits, so that more significant data can be obtained

from each subject in the time available. Furthermore, since the method gives the

threshold of acceptability directly it avoids the need for statistical analyses required

by the method of limits.

Details of the tests

Having decided in the previous section on the basic approach, in this section the

following details are described:

selection of sounds

choice of subjects

length of tests

listening room set-up

calibration of listening room

audiometric tests

test procedure.

Selection of sounds

The following options were available:

sounds from field recordings

synthesised sounds

a combination of real and synthesised sounds.

The advantage of real sounds is that they are more easily accepted as realistic. The

advantage of synthesised sounds is that they can be controlled so that only one aspect

of the sound is varied at once. Specifically, this would allow us to control the amount

of fluctuation whilst keeping other characteristics of the sound constant. The final set

presented to subjects comprised a combination of real and synthesised sounds which

was developed and refined during a series of preliminary tests.



It was decided to use at least some sounds from the field studies for realism. However,

it was not sensible to compare sounds from different case studies because a number of

factors vary, such as the frequency and character of the sounds. In order to isolate the

effects of fluctuations we needed to compare subject reactions to a number of sounds

in which all parameters (tonality, frequency content etc.) were kept constant except

the amount of fluctuation. After some searching we found a set of sounds that met this

requirement fairly closely. In case 20, the suspected source was about a mile away, so

that the fluctuation in the sound varied with wind and other factors. We were able to

select a number of short recordings from the five-day record in which the source was

essentially the same, but the degree of fluctuation of the sound varied. From this set,

the best five samples were chosen by a combination of analysis and preliminary

listening room tests. In fact there was some variation in frequency content between

the samples, but since this was not detectable by ear it was decided that they could be

considered essentially the same except for the type and strength of fluctuation. This

allowed us to combine the realism of actual sounds with the controlled fluctuations

that would otherwise have had to be synthesised. The test sounds were therefore

strongest in the 40Hz third octave band.

The sounds had to be carefully prepared. Segments of a few minutes with varying

degrees of fluctuation were identified by evaluating the standard deviation of the

sound pressure level (this is a measure of the variance in sound level, see later) over

the three nights of recording. It was verified that the sounds were „clean‟, i.e. with the

industrial source only and without extraneous noise, such as traffic, which could have

confused the picture. From this set, a smaller set of five sounds was selected in which

the frequency of the sounds was as close as possible. In preliminary tests it was found

that most of the sounds drifted in level over a period of a minute or so, making it

difficult to establish a proper threshold. Hence, a ten second sample of each sound

was taken and „looped‟ so as to produce a recording of 3 minutes duration but with a

homogenous content throughout. The „joins‟ between the looped segments were

disguised by cross fade techniques so that even expert listeners could not tell that it

had been looped.

Waveforms of the five real sounds are shown in Figure 15*.

* The sounds can be heard on http://www.acoustics.salford.ac.uk/lfn.htm



0 1 2 3 4 5 6 7 8 9 10-1

0

1

0 1 2 3 4 5 6 7 8 9 10-1

0

1

0 1 2 3 4 5 6 7 8 9 10-1

0

1

0 1 2 3 4 5 6 7 8 9 10-1

0

1

0 1 2 3 4 5 6 7 8 9 10-0.5

0

0.5

1

time, s

Figure 15: Waveforms of the real sounds Tracks 1-5 used in the laboratory tests

These real sounds did not fully answer all our needs, because it would provide results

only at a single frequency. In order to test the threshold of acceptability tests were

needed at controlled frequencies over the whole of the low frequency range. Field

recordings could not be used for this purpose because the „problem‟ frequencies

recorded on site lay in a narrow range. Also, it would not have been possible to

produce a set of sounds in which only the frequency varied in a controlled way.

Hence, the real sounds above were supplemented by synthesised sounds.

Ideally, we would have produced sounds over a range of frequencies and with a range

of fluctuation. However, this would have required too many sounds for subjects to

evaluate in the time available. Hence, two sets of synthesised sounds were used:

a set of pure tones at third octave band centre frequencies between 25Hz and

160Hz so as to cover the entire frequency range for testing the shape of the

reference curve

Sound p

ress

ure

, ar

bit

rary

unit

s



two pairs of sounds, one fluctuating at one and a half beats per second (a

beating tone) and one steady (a pure tone) so as to evaluate fluctuations at 40

and 60Hz

The „beating tones‟ were synthesised by combining two steady tones of similar

frequencies as shown in Table 5. The result was the waveforms as shown in Figure

16: Waveforms for beating tones at 40 and 50Hz used in the laboratory tests. The

frequencies of 40Hz and 60Hz were chosen because these were frequencies at which

problems occurred most often in the field studies.

40Hz beating tone 60Hz beating tone

Formed from two tones:

40Hz at 0dB

41.5Hz at –8dB

Formed from two tones:

60Hz at 0dB

61.5Hz at –8dB

Table 5: Details of how the beating tones were synthesised

0 1 2 3 4 5 6 7 8 9 10-1

-0.5

0

0.5

1

0 1 2 3 4 5 6 7 8 9 10-1

-0.5

0

0.5

1

Figure 16: Waveforms for beating tones at 40 and 50Hz used in the laboratory tests

To summarise, three sets of sounds were used:

a. Real sounds

b. Steady tones

c. Beating tones.

Choice of subjects

The choice of both the number and make up of subjects is an important consideration.

The total number of subjects was set at 18. A slightly higher number (22 subjects) was

used in a similar test in Denmark [Po03]. However, these were mostly young subjects,

and by selecting subjects with age and sex profile of sufferers the significance of the

results could be increased.



Regarding the profile of subjects, low frequency noise sufferers tend to be middle

aged or elderly, and the majority are women. Also, there is evidence that people

known to be disturbed by low frequency noise will judge sounds differently to a cross

section of non-sufferers [Pe03]. Consequently the following profile was proposed:

Group 0 3 subjects known to be disturbed by low frequency sounds

Group 1 8 subjects with the age profile of typical sufferers (55-70 year old) but

without a history of disturbance by low frequency noise

Group 2 7 subjects from a younger age group chosen at random.

Subject Age Sex Group

1 75 F 0

2 20 F 2

3 63 M 1

4 47 F 0

5 57 M 1

6 40 F 2

7 23 M 2

8 65 M 1

9 59 F 1

10 60 F 1

11 34 F 2

12 44 F 2

13 25 M 2

14 58 M 1

15 63 F 0

16 60 M 1

17 60 F 1

18 39 F 2

Table 6: Make up of subjects for laboratory test

Group Average age Sex Total

Group 0 62 3F 3

Group 1 60 5M, 3F 8

Group 2 32 2M, 5F 7

All 50 7M, 11F 18

Table 7: Make up of subject for laboratory tests by group

Length of the tests

The length of the tests was determined by the following argument. From experience

of similar tests the maximum test period over which subjects can maintain

concentration is 20 minutes after which a break is required. Also, it was considered

that a three-hour session was the maximum over which reasonable results could be

obtained from the point of view of subject fatigue. A period of training was required

of about 20 minutes, and two audiometric tests taking about half an hour in total.

Taking these constraints into account the maximum number of test session was three



with a total listening time of 1 hour. This placed a limit on the number of sounds that

could be played.

Listening room test setup

Tests were carried out in the listening room at Salford University, which conforms to

the stringent requirements of ITU-RBS1116 (standard for listening rooms). The room

is designed for comfortable listening conditions. Subjects were asked to sit on a

reclining chair, which was selected both to minimise its effect on the sound field, and

to provide a relaxing, reclined position for the night time tests.

The sound was produced through a single REL Strata V subwoofer, combined with

Genelec speakers mounted so as to give the subject the impression of being

surrounded by sound as in a real situation. Speakers were hidden from the subject by

cloth screens. Experience showed that the source could not be located by ear. The test

arrangement is shown in Figure 17.

Figure 17: Listening room setup

Once subjects were seated in the reclining chair they were read the following

instructions for the „day time‟ tests:

“Imagine you are at home during the day. Press the button whenever you consider the

sound is not acceptable to live with and keep it pressed. Whenever you consider the

sound is acceptable to live with, release the button.”

or alternatively, for the „night time‟ tests:

“Imagine you are at home at night and trying to get to sleep. Press the button

whenever you consider the sound is not acceptable to live with and keep it pressed.

Whenever you consider the sound is acceptable to live with, release the button.”

For the „day time‟ tests the main lights were on in the room, and for the „night time‟

tests the main lights were switched off leaving a low level lamp.

Reclining

chair

Operator‟s

desk

Subwoofer

Mid-range

speakers

Curtain



An operator adjusted levels using similar techniques to those used in audiometry, i.e.

by reducing the level when the button was pressed until it was released. A coarse

adjustment was made up and down to find an approximate threshold during the first

few seconds followed by finer adjustments. The operator was experienced in

audiometric testing, which helped to improve the quality of the results. The bus level

on the mixing desk was noted after each sound, and this was later calibrated to give

the sound pressure present at the ear of each subject. Each sample lasted 90 seconds,

which had been found during preliminary tests to be sufficient time to obtain a

reliable threshold. It was found by experience that, after an initial training period, the

threshold levels were repeatedly set to within 1dB, which is extremely close for this

type of test. This gave considerable confidence in the technique.

Calibration of the listening room

The listening room is specially designed to have a „flat‟ frequency response, meaning

that it is has no acoustic character that would „colour‟ the sound, for example by room

resonances. However, the usual frequency range for listening tests is down to about

40Hz, whereas in this case low frequency measurements were needed down to 25Hz

(the lowest „problem‟ frequency from field tests was 34Hz). At such low frequencies

there is no such thing as a flat response for any normal sized room. To compensate for

any colouration effects, a third octave band graphic equaliser was used and adjusted

so that the frequency response of the combined sound system and room was flat. In

fact, the sounds presented were predominantly single frequency sounds, so that

colouration effects are unlikely to play any role.

The room is also designed for low background noise. The only audible sound, apart

from the test sound was a faint buzz from amplifiers. This could have been removed,

but preliminary testing showed it to have no effect.

It was necessary to relate the bus levels, as recorded by the operator, to the actual

sound level as perceived by the subject. This was done in a calibration test before the

main block of tests in which each sound was recorded on a microphone at the position

of the subject‟s head. The sound pressure levels (Leq) recorded were used to calibrate

the bus levels.

Additionally, an overall calibration was carried out at the beginning and end of each

day, using pink noise. The variation in sound pressure level at the subject‟s head

position was never more than about ±0.2dB over the entire test run, which is within

the tolerance allowed for precision sound level meters.

Audiometric tests

There were two parts to the audiometric testing, a conventional test and a low

frequency test.

The conventional test was conducted using a Bekesy automated audiometer over the

frequency range 250Hz-6kHz. These frequencies are all above those of the sounds

presented in the listening tests, but it was considered wise to carry out the test so as to



show up any hearing defects that could affect the results. The results of these tests are

not reported here.

The low frequency audiometric tests were carried out in the anechoic chamber at

Salford University. This facility is calibrated and accredited by UKAS for testing

according to British and European standard number BSEN24869. The standard test

procedures had to be extended and modified for the purposes of this study. Firstly, the

frequency range was extended down to 31.5Hz. Secondly, pure tones were used as a

test signal rather than filtered pink noise because this is more representative of how

low frequency noise typically occurs in the field. Test frequencies were the third

octave band centre frequencies between 31.5 and 160Hz. These low frequency

hearing thresholds were needed for interpretation of subjective responses, because

individual sensitivity will affect perception.

Test procedure

All subjects participated in five separate tests

1. training period

2. audiometric testing

3. steady tones night time

4. real sounds/ beating tones day time

5. real sounds/ beating tones night time

The order of the first two sessions was reversed for half the subjects to allow access to

the audiometric facility. The order of the day and night sessions was varied randomly

in order to prevent any bias in the results.

During the training period subjects were introduced to the listening room and the

sequence of testing was explained. They were played a selection of sounds and given

some practice in adjustment of the levels. Such training periods are a widely accepted

as a necessary practice in this type of testing. In preliminary tests a training effect was

noted in which subjects tended to indicate lower thresholds the second time they were

played a sound. This was attributed to them „learning‟ to recognise the sound. Further

preliminary tests showed that a single period of training was sufficient to overcome

this effect.

Laboratory test results

Low frequency hearing thresholds

Figure 18 shows the hearing thresholds of all subjects. There is a spread of between

25 and 40dB between the most and least sensitive subjects. Figure 19 shows the

results averaged over each group. It shows that the younger age group (group 2) has

more sensitive hearing than the 55-70 year old group (group 1) by about 5dB. This

would be expected as hearing sensitivity tends to reduce with age. The shapes of the

spectra follow the published ISO values fairly faithfully, and the levels are in

agreement given that the ISO curve applies to 18-25 year olds whereas the average



age of the subjects was 60 and 32 years for group 1 and 2 respectively. (Note that the

ISO thresholds were increased by between 1 and 4 dB in 2003.)

Figure 18 also shows that the least sensitive group in terms of hearing threshold is

group 0 (sufferers). This contradicts the view sometimes expressed that those who

suffer from low frequency noise have especially acute hearing at low frequency,

although the number of subjects is too small to draw a general conclusion on this.

Low frequency hearing thresholds for all subjects

0

10

20

30

40

50

60

70

80

31.5 40 50 63 80 100 125 160


So

un

d p

ressu

re level, L

eq

, dB

Figure 18

Average low frequency hearing thresholds for each group

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

31.5 40 50 63 80 100 125 160


So

un

d p

ressu

re le

vel, L

eq

, d

B

Group 0 average Group 1 average Group 2 average ISO

Figure 19

Sufferers



Threshold of acceptability for pure tones

Figure 20 shows the thresholds of acceptability set by all subjects to tones plotted

against the frequency of the tone. There is a range of about 30dB between the most

and least sensitive subject. This is not surprising given that the thresholds of hearing

have a similar spread. Figure 21 shows the values averaged out over each group. It

shows that in absolute terms the sufferers are the least sensitive group, followed by

the older and then the younger group. As mentioned above, this contradicts the often-

held view that sufferers tend to be particularly sensitive.

Night time acceptability thresholds for tones: all subjects

0

20

40

60

80

100

120

1601251008063504031.525

Third octave band centre frequency, HzS

ou

nd

pre

ssu

re level, L

eq

, d

B

Figure 20

Night time acceptability thresholds for tones: by group

40.0

45.0

50.0

55.0

60.0

65.0

70.0

75.0

80.0

85.0

90.0

1601251008063504031.525


So

un

d p

ressu

re le

vel, L

eq

, d

B

Group 0 average Group 1 average Group 2 average



Figure 21

We would expect each individual‟s threshold of hearing to have a strong effect on

where they set the threshold of acceptability. Therefore it is interesting to see how far

above the hearing thresholds subjects set their threshold of tolerance. Shown in Figure

22 are the „relative‟ thresholds, i.e. the difference between the threshold of

acceptability and of hearing for each individual. There is about a 35dB spread in the

results. Some subjects set the threshold of acceptability only a few dB above their

hearing threshold, in other words they judged a sound that was only just audible to be

unacceptable. (In one case the threshold of acceptability is set slightly below the

threshold of hearing, which can be attributed to subject variability). Others set the

difference very much higher, so that the sound would be clearly audible before they

judged it unacceptable.

Night time acceptability thresholds for tones relative to hearing

thresholds, all subjects

-10

-5

0

5

10

15

20

25

30

35

40

45

31.5 40 50 63 80 100 125 160 Hz


Level

dif

fere

nce,

dB

Figure 22



Night time acceptability thresholds for tones relative to hearing

thresholds: by group

0.0

5.0

10.0

15.0

20.0

25.0

30.0

31.5 40 50 63 80 100 125 160


Level

dif

fere

nce,

dB


Figure 23

Figure 23 shows the values averaged by group. Two points of interest come out of

this. Firstly, there is a marked difference in the average response of sufferers

compared with the other two groups. They set the acceptable level about 10dB higher

than hearing threshold on average, whereas for non-sufferers, the difference was about

20dB. Thus, we can say that the sufferers are more sensitive in relative terms than

others (meaning relative to their hearing threshold), although as stated above in

absolute terms they were less sensitive. (Again, we should be cautious about drawing

general conclusions based on three subjects.)

The second point from Figure 23 is that the threshold of acceptability reduces, i.e gets

closer to the threshold of hearing for the lower frequency bands. For groups 1 and 2

the relative threshold in the 31.5 and 40Hz bands are lower by about 10dB than those

for higher bands. For group 0, they are lower by about 10dB between the 31.5 and

63Hz bands. (The 160Hz band is also lower, but we believe this may be an artefact of

the crossover to mid-range speakers from the subwoofer at this frequency rather than

a real effect). This is significant because it suggests that the optimum shape of a

reference curve does not follow the threshold of audibility over the whole of the low

frequency range. Rather, it will tend to follow the hearing threshold for the lower

bands but then move away from it above around 50Hz.



Relative level of national criteria compared with ISO hearing

threshold

-10

-5

0

5

10

15

20

25

20 25 31.5 40 50 63 80 100 125 160 200 250


Level

dif

fere

nce,

dB

Germany Denmark

Sweden Netherlands

Acceptability, average for all subjects

Figure 24

The thick line in Figure 24 is the threshold of acceptability relative to hearing

threshold averaged over all subjects. Shown on the same plot, are the national criteria

referenced to the ISO hearing threshold [IS03]. (Note that the ISO thresholds were

republished in 2003, with values between 1 and 4dB higher than the previous values.)

In other words all curves give the amount above or below a relevant hearing

threshold. The purpose of the plot is to compare the shapes of the curves. It can be

argued that the acceptability threshold is most similar to the Swedish curve, in that it

is flat for the lower bands and then rises, although the Swedish curve rises faster.

Threshold of acceptability for real sounds

The thresholds of acceptability for the real sounds are shown in Figure 25 for all

subjects in the „night time‟ situation. (Waveforms of these sounds are plotted in

Figure 15). There is a wide spread of results as was found for tones. This might be

expected given the wide range of hearing thresholds. However, the lines are

surprisingly parallel, which shows that all subjects responded in a similar way to the

various sounds, but at a different overall level.

Figure 26 shows the same data as Figure 25, but averaged by group. We see that, as

for the tones, group 0 is less sensitive in absolute terms than the other two groups, by

about 2dB. There is no significant difference in the responses of the other two groups.

Subjects were generally more tolerant of track 1 (which displayed the smallest

fluctuations) by about 5dB, and judged the other four sounds to be similar in terms of

their acceptability.



Night time thresholds of acceptability to real sounds: all subjects

50

55

60

65

70

75

80

85

1 2 3 4 5

Track number

Sou

nd

pre

ssure

level, L

eq, dB

Figure 25

Average night time thresholds of acceptability to real sounds: by

group

60.0

62.0

64.0

66.0

68.0

70.0

72.0

74.0

1 2 3 4 5

Track number

So

und

pre

ssu

re leve

l, Le

q, dB


Figure 26

Figure 27 and Figure 28 show the same data as Figure 25 and Figure 26, but for „day‟

rather than „night‟, and show similar trends. The average day and night curves are

shown together in Figure 29 which shows that on average respondents set the night

time thresholds 2dB lower than for the day. More importantly for this study is the fact

that the difference between day and night was almost identical for each sound, which

gives some confidence that there is not a qualitative difference in the sounds, with

some being relatively more disturbing at night.



Day time thresholds of acceptability to real sounds: all subjects

50

55

60

65

70

75

80

85

90

1 2 3 4 5

Track number

So

und

pre

ssu

re leve

l, Leq

, dB

Figure 27

Average day time thresholds of acceptability to real sounds: by

group

63.0

64.0

65.0

66.0

67.0

68.0

69.0

70.0

71.0

72.0

73.0

1 2 3 4 5

Track number

So

un

d p

ressu

re level, L

eq

, d

B


Figure 28



Comparison of average day and night time thresholds of

acceptability to real sounds

62.0

63.0

64.0

65.0

66.0

67.0

68.0

69.0

70.0

71.0

72.0

1 2 3 4 5

Track number

So

und

pre

ssu

re le

ve

l, L

eq

, dB

Day

Night

Figure 29

We would expect the acceptability thresholds set to depend on the hearing thresholds.

Therefore, as for the tones, it is useful to look at the difference between these two

thresholds for each subject. These figures are given in Figure 30 for night time and

Figure 31 for day time. Two interesting points come out of these figures:

sufferers tend to set acceptable levels very close to their threshold of hearing,

both day and night

the youngest group was most tolerant, and the older group less so to these

sounds.

Night time acceptability thresholds relative to hearing threshold for

real sounds: by group

0.0

5.0

10.0

15.0

20.0

25.0

1 2 3 4 5

Track number

Level

dif

fere

nce,

dB


Figure 30



Day time acceptability thresholds relative to hearing threshold for real

sounds: by group

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

1 2 3 4 5

Track number

Level

dif

fere

nce,

dB


Figure 31

Threshold of acceptability for ‘beating’ tones

For the beating tones, only the relative thresholds are shown for simplicity. In Figure

32 and Figure 33 are shown the night and daytime thresholds respectively, averaged

by group. There are several clear trends.

Firstly, as before, Group 0 (sufferers) is the most sensitive group in relative terms,

setting the acceptability threshold only 2-3dB above audibility threshold for night

time beating tones. Secondly, subjects were more tolerant of the steady tones than of

the corresponding beating tone by 3-5dB. Thirdly, daytime levels were set an average

of 3-4dB higher than the corresponding night time levels. Lastly, the effect of the

beating on the response was essentially the same for day and night. These last two

points are emphasised further in Figure 34.

The question arises, does Group 0 set lower levels because they were more sensitive

in the first place, or is it because they have already suffered prolonged exposure to

low frequency noise and have become sensitised (of course many other factors may

also be involved such as personality and expectations etc.)? This is an important

question when it comes to setting limits. If the latter is the case, then all subjects

might be expected to respond in a similar way after prolonged exposure. This would

give strong support to the idea that levels need to be set at or below the audible

thresholds in order to protect the majority of the population. However, as it is not

possible to envisage an ethical way to test this, it is unlikely that we will ever be able

to answer this question. One young sufferer (whose case was eventually not used)

reported that he thought he would be able to get used to the sound, but was surprised

and dismayed to find out that he couldn‟t. This tends to suggest that sensitisation may

be an issue in some cases at least, although no general conclusions can be drawn from

only one case.



Night time acceptability thresholds relative to hearing threshold for

beating tones: by group

0.0

5.0

10.0

15.0

20.0

25.0

40Hz steady 40Hz beating 60Hz steady 60Hz beating

Track number

Level

dif

fere

nce,

dB


Figure 32

Day time acceptability thresholds relative to hearing threshold for

beating tones: by group

0.0

5.0

10.0

15.0

20.0

25.0


Track number

Level

dif

fere

nce,

dB


Figure 33



Comparison of day and night time acceptability thresholds relative to

hearing threshold for beating tones: average of all subjects

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0


Track number

Level

dif

fere

nce,

dB

Day Night

`

Figure 34

Evaluation of fluctuations

Having quantified subjective response to fluctuating sound, in this section an

objective parameter is sought that reflects the responses.

Fluctuation strength

The first parameter investigated was the „fluctuation strength‟ [Te68]. This is a

relatively sophisticated parameter developed to provide a measure of sound

fluctuations for the vehicle industry. It is relatively difficult to evaluate, requiring an

appropriate computer programme which is only available on specialist equipment.

The parameter was evaluated for sounds from the field studies, but was found to give

no correlation with a subjective sense of fluctuation. It was concluded that although it

sounds promising, this parameter is not suitable for evaluation of fluctuations in LFN.

Standard deviation of sound pressure level

An alternative measure of the fluctuations is to look at the statistical distribution of

the sound pressure level sampled at set intervals. Figure 35 shows the probability

distribution plots from a 30 second sample of the 5 real sounds normalised to a mean

level (Leq) of 60dB. The height of each bar represents the length of time spent at a

particular sound level. The width of the distribution is a measure of the variation in

the sound. For example, Track 1 shows the least variation, the sound level varying

only by ±3dB from the mean, apart from a small „tail‟ of lower levels, whereas track 4

has a wider spread*. The spread of the results can most conveniently be described by

* Sounds can be heard on http://www.acoustics.salford.ac.uk/lfn.htm

http://www.acoustics.salford.ac.uk/lfn.htm



the difference between the statistical parameters L10-L90 (sometimes called the noise

climate). These parameters are available on most modern sound level meters. The

values for the five real sounds are shown in Table 8. Comparing with Figure 30 and

Figure 31 there seems to be some correlation with the thresholds of acceptability. In

particular, track 1 has the highest threshold of acceptability, typically 5dB higher than

the others and also the lowest value of L10-L90.

0

0.05

0.1

%

0

0.05

0.1

%

0

0.05

0.1

%

0

0.05

0.1

%

45 50 55 60 65 70 750

0.05

0.1

Level, dB

%

Figure 35: Distribution plots for sound levels for real sounds, Tracks 1-5

Track 1 Track 2 Track 3 Track 4 Track 5

L10-L90, dB 3.7 5.3 6.0 7.5 6.1

Average magnitude of

rate of change of level,

dB/s 27.5 32.3 32.8 31.6 32.2

Table 8: L10-L90 and rate of change of level for a 30 second sample of the real sounds used in the

laboratory tests

The relative thresholds of acceptability are plotted in Figure 36 against the value of

L10-L90 for the each sound. The points are the average for all subjects. Included on the

plot are the values for the five real sounds (diamonds), pure tones at 40 and 60Hz

(circles) and beating tones (squares – there are two points for beating tones at 40 and

60Hz, but they are so close together they cannot be distinguished).



Figure 36

In interpreting Figure 36 it is helpful to describe some findings from one of the

preliminary tests. Here subjects were played a sequence of beating tones with varying

degrees of fluctuation. We found that the thresholds of acceptability were set at about

the same level for the various beating tones, but that there was a clear difference of

about 5dB from those for the steady tones. Arguably, Figure 36 also displays this

trend: the most fluctuating sounds, represented by points to the right, were given a

„penalty‟ of about 5dB compared with steady sounds on the left. This penalty does not

go on increasing as the L10-L90 increases, but „bottoms out‟ above L10-L90 greater than

about 4 dB. There is a transition region where the penalty varies on a sliding scale

between 0 and 5dB (as marked in dotted lines). The overall trend can be simplified

without much loss of accuracy by ignoring this short transition range. The simplified

trend can then be described as follows:

L10-L90<4: no penalty


This is in a form that could be used by EHOs to decide whether to apply the 5dB

penalty.

‘Prominence’

Although the above looks promising, the difference L10-L90 is not a foolproof

parameter because it does not include any effect of the rate of fluctuations. The same

value of L10-L90 can be obtained for a slowly varying and a rapidly varying sound,

whereas experience suggests that they would be judged differently in terms of a

threshold of acceptability. The main purpose of this section is therefore to find a way

to distinguish between rapidly varying sounds (which should be given a penalty) and

sounds that vary sufficiently slowly that they are to all intents and purposes steady,

and which therefore should not be given a penalty.

A parameter has been investigated known as „prominence‟ [Pe01] (not to be confused

with the sound quality parameter of the same name). This has been suggested for

evaluation of impulsive sounds using the overall A weighted sound level. In its

Night time acceptability thresholds relative to hearing threshold for real sounds and

beating tones: variation with L10-L90

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

0 1 2 3 4 5 6 7 8

L10-L90, dB

Level

dif

fere

nce,

dB

5 dB



original form it is not therefore suitable for low frequency sound. However, we can

take part of the concept and adapt it for the current problem, namely the idea of

assessing the rate of change of the rms Fast* sound pressure level. (In fact the idea of

using the rate of change of level has been around since at least the 1970s [Ja78].) In

the method the start of an impulse is defined when the sound pressure level starts to

vary by more than 10dB per second. We would like to establish whether this is an

appropriate figure for our purposes.

Figure 37 shows the rms fast sound pressure level for a 30 second sample of the real

sounds used in the laboratory plotted instant by instant. The time-averaged rate of

change of level is given in Table 8. The sound level varies by considerably more than

10dB per second. This was true also for the beating tones. Thus, all the sounds used in

the laboratory tests, except for the steady tones exceeded the 10dB/s value and would

be classed as containing impulses according to the prominence method. However, for

slowly varying sounds the 10dB/s value would not be exceeded. On the basis of these

results then, the figure of 10dB/s seems suitable for the current purposes.

Consequently, it is suggested that a sound only be considered to be fluctuating if the

slope of the sound level (rms fast) curve exceeds 10 dB/s. Therefore, fluctuating

sounds would attract a 5 dB penalty if the value of L10-L90 exceeds 4 dB and the slope

exceeds 10 dB/s.

50

60

70

50

60

70

50

60

70

50

60

70

0 5 10 15 20 25 3050

60

70

Time, s

Figure 37: Rms Fast sound level for a 30 second sample of real sounds

* rms Fast: is the usual setting on a sound level meter for environmental noise measurement. Rms

stands for „root mean square‟, and „Fast‟, as opposed to „Slow‟ refers to the averaging time.



Conclusions from laboratory tests

In absolute terms, the sufferers in these tests were the least sensitive group to low

frequency sounds. A major factor in this is that their thresholds of hearing were higher

than other groups. We should avoid strong general conclusions because only three

sufferers were tested, and there was variation between them. Nevertheless, this

finding contradicts the view sometimes expressed that LFN problems are a result of

exceptional sensitivity.

In relative terms, sufferers tend to set the threshold of acceptability much closer to the

threshold of hearing than other groups. Whether this is because they are naturally less

tolerant, or have become sensitised by exposure is not known and probably never will

be. However, if as we suspect, it is at least in part due to sensitisation, then we would

expect all groups to respond similarly were they exposed for an extended period. This

supports the setting of levels close to the threshold of audibility.

The shape of the threshold of acceptability follows that of the hearing threshold up to

between 50 and 80 Hz and then rises. This is consistent in principle with the various

national criteria, and most closely resembles the Swedish curve.

Thresholds of acceptability were set typically 4-5dB higher for sounds with strong

fluctuations than for steady sounds. This is consistent with the Danish standard

method of adding a 5dB penalty for impulsive noise, as well as existing UK

guidelines for other types of noise (not low frequency) where a 5dB penalty is added

for noise with noticeable features. It is also consistent with previous published

research [Br94]. Therefore, we conclude:

it is appropriate to penalise fluctuating sounds compared with steady sounds

5dB is an appropriate level for any such „fluctuation penalty‟.

Fluctuation strength is not successful at quantifying low frequency fluctuations. The

most successful parameter was found to be the difference L10-L90 which has the

additional advantage that it is generally available to EHOs. Results suggest that a

penalty for fluctuations is appropriate when this value exceeds 4dB. In addition, a

sound should only be considered fluctuating when the rate of change of the rms fast

sound level in the third octave band of interest exceeds 10dB per second. The rate of

change of level is not a standard parameter, but EHOs with a PC-based, logging sound

level meter should be able to calculate the value without much difficulty. Those

without such a facility should be able to make a reasonable estimate „by eye‟ (see

Proposed Criteria for suggestions as to how this can be done).

Night time thresholds of acceptability were set 2-3dB lower than the corresponding

day time limit. This is a slightly lower difference than the 5dB day-time relaxation

used in the German standard. However, it is likely that, if anything, this difference is

underestimated in the laboratory tests, (see [In00]), so the figure of 5dB is an

appropriate amount by which to relax the limits for sounds only present during the

day.

There was consistency in the effect of fluctuations for day and night. Therefore, the

procedure used to assess fluctuations can be applied equally to night and day.



CONCLUDING REMARKS Field studies show that the various national criteria were reasonably successful in

differentiating between „positively identified‟ and „unidentified‟ problems.

Furthermore, the former correspond to cases where EHO intervention is likely to be

beneficial, and the latter to cases where they will not be able to help. On this basis,

some criteria along the lines of the various national guidelines would be of

considerable benefit to EHOs in the UK faced with complaints about LFN.

The question then arises as to which of the methods is most suitable. There is not

much evidence from the field studies to distinguish one method from another, all

worked about equally well. This is probably because the problem frequencies lie in

the frequency range where the various national reference curves are in close

agreement. However, outside this range there are some differences so the choice of

curve is important. From the results of the laboratory test the Swedish curve was

identified as being the best overall shape. Poulsen [Po03] also identified the Swedish

method as the best after the Danish one, although there was little to separate them

except for music noise, which is excluded from the scope of this study. There is an

additional advantage of the Swedish method over the Danish one in that it is

considerably easier to apply.

Comparing the Swedish curve to the German one, it is less stringent at 63Hz and

above. This is seen as a positive thing because these bands often include traffic noise

above the audible threshold. This was the case in several of the field studies where the

audible threshold was exceeded in the 80 and 100Hz band but no source of LFN was

identified. The Swedish method is therefore less likely than the German method to

falsely identify traffic noise as LFN. Thus, for several reasons, the Swedish curve is

preferred at this stage.

However, the Swedish curve only extends to 31.5Hz, whereas other methods include

frequencies down to 10Hz. There is no particular evidence from the field studies to

suggest the range should be extended below 31.5Hz. However, other experience from

the literature suggests that, although rare, problems do occasionally occur below

31.5Hz and are no less serious than above 31.5Hz. Therefore, we propose that the

Swedish curve is extended down to 10Hz. The Swedish and German curves are just

under 5dB below the ISO 226 (2003) threshold at 31.5Hz, and it is proposed to

continue this trend down to lower frequencies. ISO 226 does not give values below

20Hz, but the thresholds published by Watanbe and Moller [Wa90] can be used.

Regarding the maximum frequency, there is little evidence from the case studies to

suggest up to what frequencies should be included. On the basis of experience, the

upper frequency limit will be set at 160Hz, consistent with the Danish method.

This proposed reference curve turns out to be similar to one proposed in the

Netherlands for setting low frequency noise limits in planning applications [Sl01]. It

appears that this method is already in use on a trial basis, and is thought to work well

by those who use it. The fact that similar limits have been derived independently here

gives some measure of confidence.



Regarding fluctuations, there is evidence from the laboratory tests that a penalty for

fluctuating sounds of 5dB is appropriate. There is also evidence that such a penalty

should be applied when the difference L10-L90 exceeds 4 dB, and when the rate of

change of the rms Fast sound level in the third octave band under consideration

exceeds 10dB per second. What is not clear at present is how this penalty should

relate to the absolute level of the reference curve, i.e. whether 5dB should be

subtracted from the curves to make them more stringent, or whether the curves should

be considered to have the penalty already applied. From experience it seems likely

that most problem LFN sounds would attract the penalty, and on this assumption the

positive experience from around the world suggests that the national criteria are set at

the right level for fluctuating sounds. This being the case, it seems appropriate to

allow a relaxation of 5dB for steady sounds rather than to apply a penalty for

fluctuating sounds. This also agrees with the laboratory tests where steady sounds

were set typically more than 5dB above threshold by the most sensitive group, i.e.

sufferers, (although as stated before we should be careful about establishing absolute

levels from short exposure tests). Furthermore, one can argue that a fluctuating sound

with an average level 5dB below the threshold would be audible, whereas a steady

sound would not. Since the curve values at low frequency are set 5dB below threshold

this is again consistent with allowing a relaxation for steady sounds.

We do not have much evidence as to how long, or what proportion of the time a sound

needs to be present to become a problem. However, for case studies where a problem

was positively identified the sounds were clearly not „occasional‟ but were present on

a permanent if not continuous basis. It would have been relatively straightforward for

an EHO to decide this on the basis of measurement. Therefore, we do not propose

rigid rules but rather to leave it to the judgement of the EHO.

The equipment needed to apply the proposed method is a minimum of a sound level

meter with third octave bands down to 10Hz. This would be available to most local

authorities. Many are nowadays equipped for unmanned logging, and such equipment

would be an advantage. If audio recording is also available, this can improve the

confidence in the result. A simple method is proposed in the German method (dBC-

dBA>20dB) as an initial indicator that requires less sophisticated equipment.

However, there is evidence that although useful, this is not reliable, so it should not

form the basis for a decision.

We expect a reasonably high proportion of cases to remain unsolved even if a

criterion is adopted. This is indicated in the results of the field studies, half of which

were unidentified, and is a common experience in countries where criteria are in use.

However, this does not negate the value of a criterion which should provide EHOs

with a means of distinguishing cases where they should act from those where they can

do nothing to help. However, it does indicate the need for some alternative for those

sufferers not satisfied with the outcome. Currently, the only backup is through

voluntary organisations such as the Low Frequency Noise Sufferers Association, who

do good work but with very limited resources. An ideal complement to the proposed

criterion would be develop techniques by which the sufferer may acquire a degree of

control over their adverse reactions to the sound (see for example [Ba97]). This is

strongly recommended as an important area for further funded research (see also

[Le03].



It is suggested that the proposed criterion be used not as a prescriptive indicator of

nuisance, but rather in the sense of guidance to help determine whether a sound exists

that might be expected to cause disturbance. Some degree of judgement by the EHO is

both desirable and necessary in deciding whether to class the situation as a nuisance,

and is likely to remain so. One of the main reasons is that, from the control cases, it is

clear that problems do not necessarily arise when the criteria are exceeded. Indeed, we

can conjecture that genuine LFN complaints occur only in a few such cases.

Therefore, factors like local knowledge and understanding of the broader situation are

likely to remain important aspects of the assessment. It is thought that this approach is

likely to find acceptance since EHOs in the UK are accustomed to a fairly wide scope

in interpreting guidelines on noise nuisance.

Although sufferers often claim there is a vibration element to the noise it is rare to

find vibration levels above the perceptible limits ([Sl01, [Ru00]).



Proposed criteria and procedure for assessing low frequency noise

Measurement should be taken with the microphone in an unoccupied room where the

complainant says the noise is present. (Note that the person taking the measurements

may not be able to hear the sound).

Record Leq, L10 and L90 in the third octave bands between 10Hz and 160Hz.

If the Leq, taken over a time when the noise is said to be present, exceeds the values in

Table 9 it may indicate a source of LFN that could cause disturbance. The character of

the sound should be checked if possible by playing back an audio recording at

amplified level.

If the noise occurs only during the day then 5dB relaxation may be applied to all third

octave bands.

If the noise is steady then a 5dB relaxation may be applied to all third octave bands. A

noise is considered steady if either of the conditions a. or b. below is met:

a. L10-L90 < 5dB

b. the rate of change of sound pressure level (Fast time weighting) is less than

10dB per second*

where the parameters are evaluated in the third octave band which exceeds the

reference curve values (Table 9) by the greatest margin.

Table 9 Proposed reference curve

Hz 10 12.5 16 20 25 31.5 40 50 63 80 100 125 160

dB, Leq 92 87 83 74 64 56 49 43 42 40 38 36 34

* For a meter capable of storing short term Leq the rate of change is stLL /12 where

1L and 2L are subsequent values of the level and st is the time for each sample

(should be less than 0.1s). For simpler instruments it should be possible to estimate

the rate of change from the depth and speed of fluctuations judged by eye. For

example, if there are 2 fluctuations per second with a difference of 6dB from peak to

trough then the total change in a second is 24dB (two up, two down, each 6dB). The

rate of change would therefore be at least 24dB if the level changes smoothly, and

more than this if it changes irregularly or suddenly.



REFERENCES

[Ba97] Baguley DM, Beynon GJ, Thornton F. A consideration of the effect of ear

canal resonance and hearing loss upon white noise generators for tinnitus retraining

therapy. Journal of Laryngology and Otology, 111, 803-813, 1997.

[Br94] Bradley JS, Annoyance caused by constant-amplitude and amplitude-

modulated sounds containing rumble. Noise control Engineering Journal 42(6) 203-

208, 1994.

[DI97] DIN45680 (1997) Messungen und Bewertung Tieffrequenter

Gerauscheimmissionen in der Nachbarschaft. Beiblatt 1 Hinweise zur Beurteilung bei

gewerblichen Anlagen.

[In00] Inukai Y, Nakamura, N and Taya H, Unpleasantness and acceptable limits of

low frequency sound. Journal of Low Frequency Noise and Vibration, 19(3), 135-140,

2000.

[IS03] ISO226 (2003). Acoustics - Normal equal-loudness-level curves.

[Ja78] Jacobsen T, Measurement and assessment of annoyance of fluctuating noise.

Technical University of Denmark Report no 24, 1978.

[Le03] Leventhall G. A review of published research on low frequency noise and its

effects. Report for Defra, London 2003.

[Mi01] Mirowska M, Evaluation of low frequency noise in dwellings. New Polish

recommendations. Journal of low frequency noise, 20(2) 67-74, 2001.

[Mo02] Moller H and Lydolf M, A questionnaire survey of complaints of infrasound

and low frequency noise. Journal of low frequency noise, 21(2) 53-64, 2002

[Pe01] Pederson T H, Objective method for measuring the prominence of impulsive

sounds and for adjustment of LAeq. Proc. Internoise, 2001.

[Po03] Poulsen T and Mortensen F R. Laboratory evaluation of annoyance of low

frequency noise. Danish Environmental Protection Agency Working report no 1,

2002.

[Ru00] Rushforth I, An Integrated Acoustic / Microseismic approach to Monitoring

Low Frequency Noise & Vibration, Ph.D. Thesis, University of Liverpool 2000.

[Ru02] Rushforth I R, Moorhouse A T, Styles P The Effectiveness of DIN 45680 in

Resolving a Case Study of Low Frequency Noise. Journal of Low Frequency Noise,

Vibration and Active Control, 21(4)181-198, 2002.

[Sl01] Sloven P, A structured approach to LFS – complaints in the Rotterdam region

of the Netherlands. 20(2), 75-84, 2001.

[So01] Sorensen M F, Assessment of noise with low frequency line spectra – practical

cases. Journal of Low Frequency Noise, Vibration and Active Control, 20(4) 205-208,

2002.

[Te68] Terhardt E, Acoustic roughness and fluctuation strength. Acustica 20(4), 215,

1968.

[vB99a] van den Berg G P and Passchier-Vermeer W, Assessment of low frequency

noise complaints, Proc. Internoise99, 1999.

[vB99b] van den Berg G P, Case control study in low frequency sound measurements.

Proc. Internoise99, 1999.

[Wa90] Watanabe, T., and Møller, H. Hearing thresholds and equal loudness contours

in free field at frequencies below 1kHz. Jnl Low Freq Noise Vibn 9, 135-148, (1990)



ACKNOWLEDGEMENTS This project was funded by Defra, whose support is gratefully acknowledged.

We would like to thank all participants of field and lab studies, as well the EHOs who

helped to set up the field studies. These people cannot be named for confidentiality

reasons.

The laboratory tests were set up by Dr Bill Davies and were conducted by Paul

Kendrick.

We would like to acknowledge the contribution Geoff Leventhall in making helpful

suggestions for revision of the text.

We would also like to thank Hazel Guest, David Manley and Rosemary Mann for

their support.



APPENDIX: SUMMARY OF RESULTS FROM FIELD STUDIES

Results for case studies not presented in the main text are given here. The

experimental details were the same as for cases 20 and 2 discussed in the main text

except that a 10Hz high pass filter has been applied in Cases 13, 16, 19 and 19A.

Case 5

Measurement filename Times identified by respondent

Case5_4421_210000.cmg 04h40

Case5_4423_210000.cmg 23h20

Location Urban

Source Suspected industrial

Microphone position Corner of bedroom




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case5_040421_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case5_040421_210000.cmg






0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case5_040423_090000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case5_040423_090000.cmg





Case 6


Case6_4325_210000.cmg 08h50

Case6_4328_210000.cmg 03h50

Location Suburban





[ID=82] Average G1 1 Hz;(dB[2.000e-05 Pa], PWR) 46.25 42.6

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case6_040325_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case6_040325_210000.cmg




Case6_040325_210000.cmg together with Dutch (audibility) 39dB and Danish 50.2dB limits


0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case6_040328_210000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k



0

10

20

30

40

50

60

70

80

21h 22h 23h 00h 01h 02h 03h 04h 05h 06h 07h 08h 09h





Case 7


Case7_4514_134035.cmg 19h30

Case7_4514_134035.cmg 13h30

Location Suburban

Source Suspected industrial/ commercial

Microphone position Corner of downstairs back room



[ID=150] Average G1 Case7 Hz;(dB[2.000e-05 Pa], PSD) 70.00 31.9

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case7_040514_134035.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case7_040514_134035.cmg






0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case7_040514_134035.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k



Figure 55: Time history on 15/05/04 showing 80Hz 1/3 octave spectrum band from measurement




Case 8


Case8_4512_210000.cmg 03h30

Case8_4513_210000.cmg 04h00

Location Suburban


Microphone position Upstairs bedroom




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case8_040512_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case8_040512_210000.cmg






0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case8_040513_210000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k







Case 13


Case13_4405_210000.cmg 04h00

Case13_4406_210000.cmg 05h00

Location Suburban

Source Not known

Microphone position Corner of living room




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case13_040406_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case13_040406_210000.cmg






0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case13_040406_210000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 66: Mean 1/3 octave band spectrum 9m30s starting 05h00m on 07/04/04from



Case13_040406_210000.cmg together with lower Dutch (audibility) 22dB and Danish 39.1dB

limits



Case 16


Case16_4402_210000.cmg 02h30

Case16_4402_210000.cmg 04h10

Case16_4405_210000.cmg 02h33

Location Suburban

Source Not known.

Microphone position Corner of downstairs room

Notes Fridge/ aircraft




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 68: FFT of 9m30s audio record starting 02h30m on 03/04/04from measurement

Case16_040402_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 69: Mean 1/3 octave band spectrum 9m30s starting starting 02h30m on 03/04/04from






Spectrum at3m35s000150.00 20.15

0

250

100

200

50

150

200

0.0080.00 40

[ID=82] G1 1 3m35s000 50.31 11.21

0m00s000 9m26s0003m00s000 5m00s000

0

250

100

200

50

150

200

Cut at50.3125 Hz s;(dB[2.000e-05 Pa], PSD) 0m00s000 9.57

0m00s000 9m26s0003m00s000 5m00s000

0.00

80.00

20

40

60

c10m00s000 0.00 0.36

c29m26s000 250.00 3.04

|c2-c1|9m26s000 250.00 2.69

0.00 80.0040

Figure 71: Sonogram of 9m30s audio record starting 02h30m on 03/04/04from measurement

Case16_040402_210000.cmg. Illustrating suspected aircraft flyover with tones varying in

frequency for example from 240 to 150Hz. Constant tone at 70Hz is possibly a pc fan, while tone

at 50Hz is possibly due to the fridge with harmonics at 100, 150 and 200Hz.




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 72: FFT of 3m00s of audio record starting 04h10m on 03/04/04 from measurement

Case16_040402_210000.cmg.


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 73: Mean 1/3 octave band spectrum of 3m00s starting starting 04h10m on 03/04/04 from






Spectrum at0m00s00054.06 21.59

0

250

100

200

50

150

200

0.0060.00 30

[ID=84] G1 1 0m08s000 50.31 14.84

0m00s000 2m56s0001m00s000 2m00s0002m00s000

0

250

100

200

50

150

200

Cut at0 Hz s;(dB[2.000e-05 Pa], PSD) 0m00s000 13.80

0m00s000 2m56s0001m00s000 2m00s0002m00s000

0.00

60.00

20

40

c10m00s000 0.00 5.93

c22m56s000 250.00 0.58

|c2-c1|2m56s000 250.00 5.35

0.00 60.0030

Figure 75: Sonogram of 3m00s of audio record starting 04h10m on 03/04/04from measurement

Case16_040402_210000.cmg. Illustrating suspected aircraft flyover with tones varying in

frequency for example from 240 to 180Hz. Constant tone at 70Hz is possibly a pc fan, while tone

at 50Hz is possibly due to the fridge with harmonics at 100, 150 and 200Hz.




10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 76: FFT of 2m10s of audio record starting 02h33m from measurement

Case16_040405_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 77: Mean 1/3 octave band spectrum during 2m10s of flyover starting at 02h33 on 06/04/04

from measurement Case16_040405_210000.cmg



Case 18


Case18_4423_205000.cmg 22h10

Case18_4424_205000.cmg 05h10

Location Rural


Microphone position Corner of empty room




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case18_040423_205000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k








0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 82: FFT of 9m30s of audio record starting 05h10m on 25/04/04 from measurement

Case18_040424_205000.cmg.



0

10

20

30

40

50

60

70

80

44 45 46 47 48 49 50 51 52 53 54 55 56

Figure 83: Zoom FFT of 9m30s of audio record starting 05h10m on 25/04/04 from measurement

Case18_040424_205000.cmg showing peaks at 48.2, 48.8 and 49.2 Hz.


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 84: Mean 1/3 octave band spectrum starting 05h10m on 25/04/04 from measurement

Case18_040424_205000.cmg





Spectrum at0m00s00048.75 39.99

0

250

100

200

50

150

200

0.0050.00

[ID=47] G1 1 0m00s000 48.75 39.99

0m00s000 9m26s0003m00s000 5m00s000

0

250

100

200

50

150

200

Cut at48.75 Hz s;(dB[2.000e-05 Pa], PSD) 0m00s000 39.99

0m00s000 9m26s0003m00s000 5m00s000

0.00

50.00

20

30

c10m00s000 0.00 22.95

c29m26s000 250.00 -7.87

|c2-c1|9m26s000 250.00 30.83

0.00 50.00

Figure 86: Sonogram of 9m30s of audio record starting 05h10m on 25/04/04 from measurement

Case18_040424_205000.cmg. Illustrating falling tones varying in frequency for example from 200

to 40Hz. Constant tone at ~50Hz is due to the fridges with harmonics at 100, 150 and 200Hz.



Case 19


Case19_4421_09000.cmg 09h20

Case19_4422_09000.cmg 12h20

Location Urban

Source Plant





0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case19_040421_090000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case19_040421_090000.cmg





limits


0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case19_040422_090000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case19_040422_090000.cmg



limits



Case 19a


Case19a_4420_21000.cmg 23h10

Case19a_4422_09000.cmg 12h20

Location Urban

Source Plant

Microphone position Spare bedroom




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case19a_040420_210000.cmg


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case19a_040420_210000.cmg




Case19a_040420_210000.cmg together with Dutch (audibility) 33dB and Danish 46.2dB limits


0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Case19a_040424_090000.cmg




0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Case19a_040424_090000.cmg


Case19a_040424_090000.cmg together with Dutch (audibility) 33dB and Danish 46.2dB limits



Control Case 1

Measurement filename Times selected for presentation below

Control_case_1_4510_193000.cmg 03h00

Control_case_1_4510_193000.cmg 05h10

Location Suburban

Sources Motorway ~0.5km

Central heating pump




[ID=89] Average G1 Ch. 1 Hz;(dB[2.000e-05 Pa], PSD) 70.00 37.5

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Control_case_1_4510_193000.cmg. Background noise is mainly motorway at

~0.5km.

Ch. 1 [Average] Hz dB63 34.5

0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Control_case_1_4510_193000.cmg




Control_case_1_4510_193000.cmg together with Dutch (audibility) 33dB and Danish 46.2dB

limits. Noise sources are motorway and central heating.

[ID=90] Average G1 Ch. 1 Hz;(dB[2.000e-05 Pa], PSD) 35.94 62.3

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240

Figure 102: FFT of 9m30s of audio record starting 06h00m from measurement

Control_case_1_4510_193000.cmg. Background noise is mainly central heating pump.



Average G1 Ch. 1 [Average] Hz;(dB[2.000e-05 Pa], PWR) 40 57.7

0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k

Figure 103: Mean 1/3 octave band spectrum starting 06h00m from measurement

Control_case_1_4510_193000.cmg


Control_case_1_4510_193000.cmg together with Polish (audibility) 44.6dB and Danish 54.6dB

limits. Noise sources are motorway and central heating.



Control Case 3

Measurement filename Times selected for presentation below

Control_case3_4511_171944.CMG.cmg 17h19

Location Ground floor city centre flat

Sources City centre traffic

Microphone position Corner of living room




0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160 180 200 220 240


Control_case3_4511_171944.CMG.cmg.


0

10

20

30

40

50

60

70

80

1 2 4 8 16 31.5 63 125 250 500 1 k 2 k 4 k 8 k


Control_case3_4511_171944.CMG.cmg




Control_case3_4511_171944.CMG.cmg together with Dutch (audibility) 33dB and Danish 46.2dB

limits. Main noise source is city centre traffic.

proposed criteria for the assessment of low frequency

Documents